Community
Community

Exciting Announcement! In celebration of launching our AI Certification, we’re thrilled to offer a 50% discount exclusively. Seize this unique chance—don’t let it slip by!

Learn more

Category: SONiC

Categories
Open Networking Enterprise Suite SONiC

The Power of Choice in Networking: How The AI Stack Breaks Down Barriers

A lot of people ask me, “What are the problems that you are solving for customers?” At Aviz, we understand that modern networking demands more than just connectivity; it requires agile, scalable solutions that can adapt to the evolving demands of AI-driven environments. We’re tackling the challenges of complexity, vendor lock-in, and prohibitive costs that many face in traditional network setups. Our AI Networking Stack isn’t just about keeping your network running; it’s about advancing it to think, predict, and operate more efficiently.

At Aviz, we are reshaping networks for the AI era by pioneering both ‘Networks for AI’ and ‘AI for Networks’. Our AI Networking Stack offers unparalleled choice, control, and cost savings, designed to enhance orchestration, observability, and real-time alerts in a vendor-agnostic environment. We’re not just providing solutions; we’re transforming networks with advanced LLM-based learning for critical operations, ensuring powerful, open-source solutions that drive innovation at a fraction of the cost.

We lead the journey to redefine networking with a data-centric approach that seamlessly integrates with any switch and network operating system, delivering performance that rivals the top OEM solutions—all while focusing on the core pillars of choice, control, and cost-effectiveness.

So, if you value having choices, staying in control, and achieving cost savings, read on to discover how our innovative solutions can transform your network management experience.

Now, let’s take a closer look at what sets our technology apart. Here is the detailed overview of our AI Networking Stack:

We’ve meticulously developed each layer of our AI Networking Stack to address the unique challenges our customers face in today’s dynamic network environments. From foundational hardware choices to advanced AI-driven functionalities, let’s dive into the specifics of what makes our technology stand out in the industry.

First, let’s discuss why choosing a vendor-agnostic approach is so crucial. Imagine using Linux. Does it really matter you use it on HPE, DELL, Lenovo Servers, or even at AWS, Azure or GCP. That’s the kind of interoperability we bring to the networking world. Similar to what Linux did for the tech industry, we leverage the open-source SONiC operating system, enhanced by strong partnerships and robust community support. This approach offers an array of choices, enabling hardware selections from our partner ecosystem without any constraints.

At the heart of our innovative lineup is ONES (Open Networking Enterprise Suite), which empowers you with real-time visibility, seamless orchestration, advanced anomaly detection, and AI fabric functionality, including RoCE, across multiple vendors.
This means you have full control over which hardware solutions you implement, supported by our dedicated 24/7 customer service. ONES is designed to give you the freedom to manage your network without vendor lock-in, ensuring flexibility and control in your hands. Another critical aspect of our strategy to ensure cost efficiency is the Open Packet Broker (OPB).
Built on the powerful, community-driven SONiC platform, the OPB mirrors the capabilities of traditional Network Packet Brokers (NPB) but at a fraction of the cost. This solution delivers all the traditional functionalities you expect but optimizes them to offer significant cost savings without sacrificing performance or scalability.
Sitting atop our stack is the GenAI-based Network Copilot, your AI-powered assistant that simplifies all aspects of network management—from routine upgrades and audit reports to complex troubleshooting tasks. This tool is designed to enhance your operational efficiency, dramatically reducing the time and effort required for network management tasks, thereby freeing up your team to focus on strategic initiatives that drive business growth.

Our AI Networking Stack is designed to be the backbone of future network management, integrating advanced AI to navigate the complexities of modern networking with sophistication and ease. Opting for a vendor-agnostic approach provides the flexibility to choose the best technologies at the most effective prices, ensuring your network remains robust, scalable, and primed for future technological advancements.

Explore the benefits of a networking solution that brings choices, control, and cost savings without the constraints of traditional vendor lock-ins. This approach is not just about adopting new technology—it’s about advancing with a platform that understands and adapts to the evolving needs of your enterprise.

FAQs

1. What is an AI Networking Stack and why is it important for modern enterprises?

The AI Networking Stack refers to a multi-layered, AI-powered network management architecture that provides real-time orchestration, anomaly detection, and observability across multivendor environments. Aviz AI Stack combines ONES, Open Packet Broker (OPB), and Network Copilot™ to deliver flexibility, vendor-agnostic integration, and cost-efficiency—ideal for businesses modernizing their networks for AI-era workloads.

Unlike traditional OEM networks that tie users into proprietary hardware and software, Aviz’s open-source SONiC-based stack supports interoperability across Dell, HPE, Cisco, Arista, and others. This vendor-agnostic approach lets enterprises avoid lock-in, reduce costs, and adapt faster to evolving infrastructure needs.

Aviz’s stack includes:

  • ONES for orchestration, visibility, and AI fabric monitoring
  • Open Packet Broker (OPB) for cost-efficient traffic visibility
  • Network Copilot™ for AI-powered network insights and automation

These tools work together to deliver full-stack AI networking without dependency on any specific NOS or hardware.

Network Copilot™ leverages LLM-based AI to provide intelligent chat-based troubleshooting, performance diagnostics, upgrade checks, and real-time analytics. It streamlines operations and replaces repetitive tasks with intelligent automation—making it ideal for both network engineers and business leaders.

SONiC (Software for Open Networking in the Cloud) enables hardware-agnostic deployment, similar to how Linux enabled OS flexibility. Aviz builds on SONiC to offer full-stack solutions (ONES, OPB, Copilot) that allow organizations to choose hardware freely, while retaining full control over orchestration, observability, and AI integration.

Categories
Open Networking Enterprise Suite SONiC

ONE Data Lake & AWS S3 – Enhancing data Management and Analytics – Part 2

In February, we introduced the ONE Data Lake as part of our ONES 2.1 release, highlighting its integration capabilities with Splunk and AWS. In this blog post, we’ll delve into how the Data Lake integrates specifically with the S3 bucket of AWS.

A data lake functions as a centralized repository designed to store vast amounts of structured, semi-structured, and unstructured data on a large scale. These repositories are typically constructed using scalable, distributed, cloud-based storage systems such as Amazon S3, Azure Data Lake Storage, or Google Cloud Storage. A key advantage of a data lake is its ability to manage large volumes of data from various sources, providing a unified storage solution that facilitates data exploration, analytics, and informed decision-making.

Aviz ONE-Data Lake acts as a platform that enables the migration of on-premises network data to cloud storage. It includes metrics that capture operational data across the network’s control plane, data plane, system, platform, and traffic. As an enhanced version of the Aviz Open Networking Enterprise Suite (ONES), ONE-Data Lake stores the metrics previously used in ONES in the cloud.

Why AWS S3?

Amazon S3 (Simple Storage Service) is often used as a core component of a data lake architecture, where it stores structured, semi-structured, and unstructured data. This enables comprehensive data analytics and exploration across diverse data sources. S3 is widely used for several reasons:
  • 1. Effective Integration and Ecosystem:
S3 integrates seamlessly with a wide range of AWS services and third-party tools, significantly enhancing data processing, analytics, and machine learning workflows. This seamless integration allows for efficient data ingestion, real-time analytics, advanced data processing, and robust machine learning model training and deployment, creating a powerful and cohesive ecosystem for comprehensive data management and analysis.
  • 2. Durability and Reliability:
S3 is engineered for complete durability, ensuring that your data is exceptionally safe and consistently accessible. This level of durability is achieved through advanced data replication across multiple geographically dispersed locations, providing robust protection against data loss and guaranteeing high availability.
  • 3. Security:
S3 offers comprehensive security and compliance capabilities, providing a robust framework for safeguarding data and ensuring regulatory adherence. This includes advanced data encryption, both at rest and in transit, ensuring that sensitive information remains protected throughout its lifecycle. Additionally, S3 provides granular access management tools, such as AWS Identity and Access Management (IAM), bucket policies, and access control lists (ACLs), allowing fine-tuned control over who can access and modify data. These features, combined with compliance certifications for various industry standards (such as GDPR, HIPAA, and SOC), make S3 a secure and reliable choice for data storage in highly regulated environments.
  • 4. Scalability:
S3’s capability to handle virtually unlimited amounts of data makes it an unparalleled choice for building and maintaining expansive data lakes that require storing massive volumes of information. This scalability empowers organizations to seamlessly scale their storage needs without upfront investments in infrastructure, accommodating growing data demands effortlessly. This capability is crucial for enterprises seeking to centralize and manage diverse data types, enabling advanced analytics, machine learning, and other data-driven initiatives with agility and reliability.
  • 5. Cost Effectiveness:
S3 provides flexible pricing models and a variety of storage classes to optimize costs based on data access patterns. Users can take advantage of storage classes like S3 Standard, S3 Intelligent-Tiering, S3 Standard-IA, S3 One Zone-IA, and S3 Glacier to manage expenses efficiently.
  • 7. Disaster recovery and Backup:
S3 offers robust data management capabilities, including versioning, lifecycle policies, and replication, which streamline data governance and archival processes. These features ensure data integrity, compliance, and resilience across various use cases, from regulatory compliance to disaster recovery planning. This capability empowers businesses to unlock the full potential of their data assets, supporting diverse applications such as predictive analytics, business intelligence, and real-time reporting with ease and efficiency.
  • 6. Data Management and Analytics:
S3’s robust features, including cross-region replication and lifecycle policies, establish it as an exceptional solution for disaster recovery strategies, ensuring data redundancy and resilience. Furthermore, S3’s lifecycle policies enable automated management of data throughout its lifecycle, facilitating seamless transitions between storage tiers and automated deletion of outdated or unnecessary data. Together, these features make S3 a reliable backup solution that enhances data durability and availability, providing organizations with peace of mind knowing their critical data is securely stored and accessible even in unforeseen circumstances.

Integrating S3 with ONES:

Steps involved to integrate the S3 cloud service with ONES,
  • 1. Mapping S3 instance with the ONES server
To integrate the S3 service with ONES, follow these steps:
  • Configure S3 Instances: Set up the S3 instances on the ONES cloud page to start pushing metrics to the designated cloud endpoint.
  • Provide Necessary Details: The following information is required for the integration:
  • ARN Role: The unique identifier for the role that grants permissions to access specific AWS resources, including S3 buckets
  • Region: The AWS region where your S3 bucket is located
  • Bucket Name: The globally unique name of your S3 bucket
  • External ID(Optional): An external ID is an additional security measure used when granting cross-account access to IAM roles.
By accurately providing these details, you can effectively configure and integrate the S3 service with ONES, facilitating smooth metric collection and analysis.
Figure 1: Cloud Instance configuration page in ONES
Figure 2: Instance created and ready for data streaming
  • 2. Managing the created Instance through ONES:
The cloud instance created within ONES offers several management options to enhance user experience and sustainability. Users can update the integration settings, pause and resume metric uploads to the cloud, and delete the created integration when needed. These features make it easy for users to maintain and manage their cloud endpoint integrations effectively.
Figure 3 : Updating the integration details
Figure 4: Option to pause and resume the metric streaming to cloud
Figure 5: Option to delete the integration created
  • 3. User defined metric update:
The end user has the flexibility to select which metrics from their network monitored by ONES should be uploaded to the designated cloud service. This ONES 2.1 release supports various metrics, including Traffic Statistics, ASIC Capacity, Device Health, and Inventory. Administrators can choose and deselect metrics from the available list within these categories according to their preferences.
Figure 6 : Multiple options available for metric update on cloud
  • 4. Multi vendor support
The metric update is not limited to any particular hardware or network operating system (NOS). ONE-Data Lake’s data collection capability extends across various network operating systems, including Cisco NX-OS, Arista AOS, SONiC, and Non-SONiC. Data streaming occurs via the gnmi process on SONiC-supported devices and through SNMP on OS from other vendors.
Figure 7: ONES inventory showing multiple vendor devices streaming

S3 Analytical capabilities:

Analyzing data stored in an S3 bucket can be accomplished through various methods, each leveraging different AWS services and tools. Here are some key methods:

AWS Athena:

Description: A serverless interactive query service that allows you to run SQL queries directly against data stored in S3.

Use Case: Ad-hoc querying, data exploration, and reporting.

Example: Querying log files, CSVs, JSON, or Parquet files stored in S3 without setting up a database.

Figure 8 - Data stored in S3 bucket in json format

AWS Glue:

Description: A managed ETL (Extract, Transform, Load) service that helps prepare and transform data for analytics.

Use Case: Data preparation, cleaning, and transformation.

Example: Cleaning raw data stored in S3 and transforming it into a more structured format for analysis.

Figure 9 - Pie Chart in S3 representing the data from different NOS vendors

AWS SageMaker:

Description: A fully managed service for building, training, and deploying machine learning models.

Use Case: Machine learning and predictive analytics.

Example: Training machine learning models using large datasets stored in S3 and deploying them for inference.

Third-Party Tools:

Description: Numerous third-party tools integrate with S3 to provide additional analytical capabilities.

Use Case: Specialized data analysis, data science, and machine learning.

Example:Using tools like Databricks, Snowflake, or Domo to analyze and visualize data stored in S3.

Custom Applications:

Description: Developing custom applications or scripts that use AWS SDKs to interact with S3.

Use Case: Tailored data processing and analysis.

Example: Writing Python scripts using the Boto3 library to process data in S3 and generate reports.

Conclusion:

Aviz ONE-Data Lake serves as the cloud-native iteration of ONES, facilitating the storage of network data in cloud repositories. It operates agnostically across various cloud platforms and facilitates data streaming from major network device manufacturers like Dell, Mellanox, Arista, and Cisco. Network administrators retain flexibility to define which metrics are transferred to the cloud endpoint, ensuring customized control over the data storage process.

Unlock the ONE-Data Lake experience— schedule a demo on your preferred date, and let us show you how it’s done!

FAQs

1. What are the benefits of integrating ONE Data Lake with AWS S3 for network data storage?

 Integrating Aviz ONE Data Lake with AWS S3 enables:

  • Centralized cloud storage for network telemetry
  • Unlimited scalability for growing datasets
  • Enhanced security with AWS encryption and IAM controls
  • Durable and highly available storage across regions

Flexible analytics through services like AWS Athena, Glue, and SageMaker
 This combination helps enterprises achieve cost-effective, compliant, and powerful data management.

To set up AWS S3 integration with ONE Data Lake:

  • Provide your ARN role, region, bucket name, and (optionally) external ID
  • Configure your S3 instance on the ONES cloud interface

Select desired network metrics (e.g., Traffic Stats, Device Health) for uploading
 This ensures seamless cloud metric collection customized to your organization’s needs.

With ONES 2.1, administrators can selectively upload metrics like:

  • Traffic Statistics
  • ASIC Capacity Metrics
  • Device Health and Platform Monitoring
  • Inventory Data
     The flexibility to customize and filter metrics helps optimize storage costs and streamline analytics pipelines.

Yes!
 Aviz ONE Data Lake collects and streams telemetry across multiple NOS platforms, including:

  • Cisco NX-OS
  • Arista EOS
  • SONiC
  • Cumulus Linux and other non-SONiC devices .It uses gNMI for SONiC and SNMP for other vendors, ensuring multi-vendor support without limitations.

  • AWS Athena enables SQL-based querying directly on raw S3 data (no database setup needed).

  • AWS Glue automates ETL workflows, prepping raw network telemetry for structured analytics.
  • AWS SageMaker builds ML models using S3-stored datasets for predictive network optimization.

Together, these services transform raw network data into actionable insights and machine learning opportunities.

Categories
Open Networking Enterprise Suite SONiC

ONE Data Lake & Splunk: Revolutionizing Network Data Analytics – Part 1

In February, we introduced the ONE Data Lake as part of our ONES 2.1 release, highlighting its integration capabilities with Splunk and AWS. In this blog post, we’ll delve into how the Data Lake integrates specifically with Splunk.

A data lake serves as a centralized storage facility capable of accommodating large quantities of structured, semi-structured, and unstructured data on a significant scale.These are typically built using scalable distributed cloud-based storage systems, such as Amazon S3, Azure Data Lake Storage, or Google Cloud Storage.

A pivotal benefit of a data lake lies in its capacity to handle substantial amounts of data from diverse origins, offering a cohesive storage solution conducive to data exploration, analytics, and informed decision-making processes.

Aviz ONE-Data Lake functions as a platform facilitating the migration of on-premises network data to cloud storage. It encompasses metrics that capture operational data across the network’s control plane, data plane, system, platform, and traffic. Serving as an upgraded iteration of Aviz Open Networking Enterprise Suite (ONES), ONE-Data Lake stores the metrics previously utilized in ONES onto the cloud.

Why Splunk?

Splunk is highly significant for organizations across diverse industries for multiple reasons:
  • 1. Operational Insight:
Splunk empowers organizations to obtain immediate insights from their operational data, facilitating the monitoring of system and application health and performance. This capability aids in promptly identifying and addressing issues, thereby reducing downtime and enhancing operational efficiency.
  • 2. Security Surveillance and Threat Identification:
Splunk is extensively utilized for Security Information and Event Management (SIEM) objectives, aiding organizations in overseeing their IT environments for security threats and irregularities. By correlating data from diverse sources, it can efficiently identify and address security incidents, thereby bolstering the overall cybersecurity stance.
  • 3. Regulatory Adherence and Oversight:
Splunk supports regulatory adherence and oversight by empowering organizations to gather, analyze, and report on data pertinent to regulatory requirements and industry standards. This capability is especially critical for sectors like finance, healthcare, and government, where stringent compliance mandates are in place.
  • 4. IT Operations and DevOps:
Splunk aids in IT operations and DevOps practices by providing visibility into IT infrastructure, application performance, and deployment processes. This allows organizations to identify areas for optimization, streamline operations, and accelerate the development and delivery of software applications
  • 5. Machine Learning and Predictive Analytics:
Splunk equips organizations with machine learning and predictive analytics functionalities, empowering them to uncover patterns, detect anomalies, and forecast outcomes from their data. This supports proactive resolution of issues, capacity planning, and efforts in risk management
  • 6. Customer Satisfaction Administration:

Splunk can be utilized to assess customer interactions and feedback from various channels, enabling organizations to delve deeper into customer requirements and preferences. This information can then be utilized to tailor offerings, elevate customer satisfaction levels, and nurture brand loyalty

To sum up, Splunk is an essential tool for organizations to leverage data efficiently, promoting operational excellence, strengthening security measures, ensuring compliance, and achieving business objectives.

Integrating Splunk with ONES:

Steps involved to integrate the Splunk cloud service with ONES,
  • 1. Mapping Splunk instance with the ONES server
To integrate the Splunk service with ONES, follow these steps:
  • Configure Splunk Instances: Set up the Splunk instances on the ONES cloud page to start pushing metrics to the designated cloud endpoint.
  • Provide Necessary Details: The following information is required for the integration:
  • Splunk URL: The URL of your Splunk instance.
  • Unique Token: A unique token for authentication and secure data transmission.
  • Index: The specific index in Splunk where the pushed metrics will be stored.
By ensuring these details are accurately provided, you can successfully configure and integrate the Splunk service with ONES, enabling seamless metric collection and analysis.
Figure 1: Cloud Instance configuration page in ONES
Figure 2: Instance created and ready for data streaming
  • 2. Managing the created Instance through ONES:
The cloud instance created within ONES offers several management options to enhance user experience and sustainability. Users can update the integration settings, pause and resume metric uploads to the cloud, and delete the created integration when needed. These features make it easy for users to maintain and manage their cloud endpoint integrations effectively.
Figure 3 : Updating the integration details
Figure 4: Option to pause and resume the metric streaming to cloud
Figure 5: Option to delete the integration created
  • 3. User defined metric update:
The end user has the flexibility to select which metrics from their network monitored by ONES should be uploaded to the designated cloud service. This ONES 2.1 release supports various metrics, including Traffic Statistics, ASIC Capacity, Device Health, and Inventory. Administrators can choose and deselect metrics from the available list within these categories according to their preferences.
Figure 6 : Multiple options available for metric update on cloud
  • 4. Multi vendor support
The metric update is not limited to any particular hardware or network operating system (NOS). ONE-Data Lake’s data collection capability extends across various network operating systems, including Cisco NX-OS, Arista AOS, SONiC, and Non-SONiC. Data streaming occurs via the gnmi process on SONiC-supported devices and through SNMP on OS from other vendors.
Figure 7: ONES inventory showing multiple vendor devices streaming

Splunk Analytical capabilities:

  • 1. Event based Visualization:
Events within Splunk generally contain timestamped data alongside related metadata and content. Each event undergoes parsing and indexing separately, facilitating users to efficiently search, analyze, and visualize data. Splunk automatically extracts fields from events during indexing, streamlining filtering and correlation based on specific criteria.
Figure 8 - Inventory details from NX-OS is captured as events in Splunk
  • 2. Chart representation:
This entails visually depicting data using charts or graphs, aiding users in comprehending patterns, trends, and relationships within the data more readily than analyzing raw data alone. These graphical representations encompass diverse types such as bar charts, line charts, pie charts, scatter plots, and others, each tailored to specific data types and analytical objectives
Figure 9 - Pie Chart in Splunk representing the data from different NOS vendors

Conclusion:

Aviz ONE-Data Lake functions as the cloud-based version of ONES, enabling the storage of network data in cloud repositories. It operates independently of any particular cloud platform and supports data streaming from leading network device manufacturers such as Dell, Mellanox, Arista, and Cisco. Network administrators have the freedom to specify the metrics they want to transfer to the cloud endpoint, granting customized control over the data storage procedure.

Schedule your demo today because with ONE Data Lake integrated with Splunk, you’re not just managing data — you’re revolutionizing network analytics for unparalleled insights and efficiency.

FAQs

1. What is Aviz ONE Data Lake and how does it enhance network data analytics?

 Aviz ONE Data Lake is a cloud platform that collects and stores telemetry from multi-vendor networks.It centralizes operational, traffic, and device health data — giving teams a single source for deep analytics, proactive decisions, and smarter network management.

Connecting ONE Data Lake with Splunk gives teams:

  • Real-time analytics
  • Powerful dashboards
  • Anomaly detection
     It helps detect issues faster, optimize resources, strengthen security, and improve operational visibility across all network layers.

You can stream:

  • Traffic statistics
  • ASIC utilization
  • Device health
  • Network inventory
     from SONiC and non-SONiC devices like Cisco, Arista, Dell — using gNMI and SNMP protocols.

No!
 ONE Data Lake is vendor-neutral.
 It supports multi-vendor environments — including SONiC, Cisco NX-OS, Arista EOS, and more — so you get unified observability across your entire network.

Through ONES, users can:

  • Update integration settings
  • Pause or resume uploads
  • Select which metrics to send

Delete integrations when needed
 It’s simple, flexible, and designed for dynamic network environments.

Categories
Open Networking Enterprise Suite SONiC

Pioneering Networking in the AI Era: Collaboration with Aviz and Marvell

At Aviz Networks, we’re dedicated to revolutionizing AI networking solutions by introducing AI for Networks and Networks for AI. Our strategic collaboration with Marvell, a global leader in semiconductor innovation, marks a significant step towards driving the adoption of SONiC. This partnership leverages Marvell’s cutting-edge switch technology and our Open Network Enterprise Suite (ONES) to deliver advanced AI Fabric capabilities and SONiC solutions for edge deployments. By integrating these technologies, we address the increasing demands of AI-driven applications, catering to both core and edge network environments with tailored solutions.

SONiC for AI Fabric and Edge: Distinct Use Cases

Our collaboration with Marvell focuses on two distinct use cases: SONiC for AI fabric and for edge deployments. Each serves a unique purpose in the evolving landscape of AI-driven networking.

SONiC for AI Fabrics

The combination of Marvell’s switch technology and Aviz Networks’ ONES platform provides a robust solution for core network environments. This integrated approach delivers:
  • Enhanced Performance: Superior performance for AI-driven applications, ensuring efficient data handling and reduced latency. This is powered by ONES' unique capability of Orchestration and Observability for RoCE traffic.
  • Scalability: Scalable networking infrastructure capable of supporting the growing demands of AI workloads from edge to cloud environments. ONES' horizontally scalable design ensures that no matter how big the AI Fabric grows, network operators can manage it from a single pane of glass.
  • Reliability: Advanced real-time visibility and anomaly detection features enhance network reliability and ensure uninterrupted AI operations. ONES' RoCE-specific anomaly detection ensures the network reliably delivers packets end to end.

SONiC for Edge Deployments

The PENS (PoE Edge Networks with SONiC) workgroup, formed in collaboration with Aviz Networks, the Linux Foundation, Marvell, and others, adapts SONiC for edge LANs. This initiative integrates specialized protocols essential for enhancing connectivity and network efficiency at the enterprise edge. Key benefits include:
  • Localized AI Processing: By deploying SONiC solutions at the edge, enterprises can process AI workloads locally, reducing latency and improving response times for real-time applications.
  • Cost Efficiency: Minimizing the need for extensive data transport to centralized cloud locations results in cost savings on bandwidth and improves overall efficiency.
  • Enhanced Security: Processing data at the edge enhances security by keeping sensitive information local and reducing the risk of data breaches during transmission.

Aviz Networks and Marvell: A Best-of-Breed Disaggregated Offering

The collaboration between Marvell and Aviz Networks provides cloud and enterprise data centers with a SONiC-based solution using Marvell switch silicon. Marvell delivers switch silicon with SAI (Switch Abstraction Interface) to integrate seamlessly with open-source SONiC, while we at Aviz Networks offer disaggregated support and services alongside cloud-native applications. Aviz Networks’ ONES (Open Networking Enterprise Suite) stack provides orchestration, telemetry, assurance, and support, ensuring that key network metrics meet SLAs and offering insights into real-time network health.

Successful Deployment and Future Prospects

Together, Marvell and Aviz Networks have successfully deployed hundreds of 400G fabric switches with SONiC running on Marvell silicon in enterprise data centers. These leaf and spine switches, equipped with 100G and 400G ports, deliver industry-leading innovation, a choice of optics, and broad skill-set availability, resulting in significant OPEX and CAPEX savings. We anticipate extending this successful collaboration to many more enterprise and cloud data center operators as the benefits of disaggregated networking become more widely understood and embraced.

Voices from the Partnership

We are excited to collaborate with Aviz Networks to enhance AI-driven networking solutions. Our switch technology, combined with Aviz Networks' expertise in AI fabric management and SONiC for edge deployments, will provide customers with a powerful, scalable, and reliable network infrastructure.

Rishi Chugh, Vice president of product marketing, Network Switching Business Unit, at Marvell

Our partnership with Marvell is a testament to our dedication to advancing AI-driven networking solutions. By combining our innovative ONES platform with Marvell's leading switch technology, we are poised to deliver unparalleled performance and reliability for our customers, both at the core and the edge.

Vishal Shukla, Founder and CEO of Aviz Networks
Categories
Open Networking Enterprise Suite SONiC

My Unique Learnings with Aviz Networks: Developing ONES Stack to Empower SONiC Adaptation & Supportability

Today, the open source and network communities recognize SONiC as the next ‘big thing’ to accelerate innovation and growth. And that’s undoubtedly true. SONiC is fundamental in building next-generation networks. However, deploying and monitoring a SONiC network is not a cakewalk that simply requires a “download, install and complete” approach.

Through this article, I, Krupakar Annam, would like to share my journey with SONiC, and how it can empower enterprises across the globe.

With a burning passion for innovation and a drive to make a real impact, I eagerly took on the challenge of leading the development team. And let me tell you, it has been an exhilarating journey. What sets our product ideas apart is that we’re not building yet another SONiC Distro or a yet another NetOps solution. In fact, making a Distro has never been a part of our plan. Our main focus is to enable customers to migrate and manage SONiC at a reasonable cost along with a lightning-fast speed. It’s a win-win situation for them, isn’t it?

Pioneering the Future of Networking with ONES

As I delved deeper into the world of SONiC, I became more confident about its untapped opportunities to transform the networking landscape. Most importantly, the power of open-source, the flexibility of disaggregated networks, and the promise of cost-effective scalability—all aligned perfectly with Aviz Networks’ vision.

And as my journey continued with this company, we went the extra mile together to understand the challenges and came up with an industry-leading solution. Here are the details:

- Pain Points of SONiC Adaptation and Management

Though SONiC is open-source, there is no single NetOps-like tool that can support multi-vendor environments. On the other hand, the available solutions are tied to specific vendors.

The pain points do not end here. There is no unified solution to handle the pre-deployment process for a multi-vendor DC fabric, whether it involves re-validating use case scenarios or creating a unified solution to bring vendors together for support.

Furthermore, there is also a lack of an all-in-one solution for multi-vendor visibility, orchestration, and supportability that makes it easier to manage DC and make more accurate decisions quickly.

We also realized many customers who are using SONiC or willing to adopt it, ultimately end up with ad hoc solutions by putting together in-house or open-source tools. These solutions are bound by specific use cases and may have limitations in pulling in-depth metrics offered by SONiC—leading to a painful process while transforming the DC fabrics.

- The Game Changer Solution

They say it right, understanding the challenges is the key to better innovations. And that’s what drove us to create ONES (Open Networking Enterprise Suite).

It’s a comprehensive, flexible, and unified stack solution. ONES can support multi-vendor SONiC platforms while adapting to or managing SONiC-based networks’ pre-deployment, orchestration, visibility, and supportability aspects of their DC network.

Our approach is centered on streamlining Day 1, and Day 2 operations of the DC network with the belief that our stack solution can run on any SONiC platform (regardless of the underlying hardware) and cater to diverse customer use cases. Whether it’s through a user-friendly GUI or robust APIs, it should be a rock-solid supportability tool for SONiC-based DC fabrics.

Challenges and Achievements in Implementing ONES Stack for SONiC Platforms and Beyond

- Unifying Network Configuration for Multi-Vendor Environment

The challenge here was to accommodate all vendor-specific configuration methods in a single go.

To address this, our solution has a common intent-based template that takes the fabric-level configurations, and the ONES orchestration layer with the intelligence to split that into vendor-specific configurations. The ONES orchestration agent running on the specific device will take care of pushing the right config to the device using CLIs and ConfigDB.

The SONiC CLI is incomplete or not a unified way of configuring and managing all available features on a device. So we have come up with FMCLI (Fabric Manager Command Line Interface), which is far more flexible and easier to adapt than SONiC CLI.

It wasn’t an easy task as we needed to decode each unique use case of many customers and include all of them without disrupting the intent of the solution we were building. What started with one or two community SONiC support has kept expanding orchestration via intent-based configuration for large networks (IPCLOS, L2 VXLAN MCLAG, L3 VXLAN MCLAG, FRR, etc.).

Furthermore, the network configuration APIs of ONES are exposed with an easily usable Python SDK to help the current infrastructure of DC adapt to ONES easily.

- Normalizing Multi-Vendor Deep Metrics for Visibility

The challenge was to pull in deep metrics at regular intervals and normalize them for unified visibility. We developed a telemetry layer that pulls 200+ metrics using gNMI and normalizes them. For deep metrics data, we have gone beyond RedisDB. Our goal is to provide maximum metrics in a presentable format to help network operators correlate and diagnose issues to make quicker decisions.

We have open sourced ONES Valdiated Designs (https://ovd.ai/), they can help to integrate ONES seemlessly into your existing infrastructure.

- Accommodating Multi-NOS Environment

We cannot expect DC fabrics to run on SONiC alone. For customers looking for migration to SONiC, they might choose to go in a phased manner. As a result, we may end up in a situation where SONiC and non-SONiC devices coexist in the same fabric. This presents a unique challenge since non-SONiC devices are proprietary, and well-established vendor-specific NetOps tools are available to manage them.

So with ONES, customers can visualize non-SONIC devices in the same fabric and limited metrics along with SONiC devices. This is achieved by using eAPIs, NVAPIs and of course, normalized metrics for visualization.

Furthermore, we have devised a distinct method to display the entire fabric in its topology while also enhancing it with alerts, rules, and other capabilities.

Transforming ONES into a Complete Deployment Solution

Our goal of supporting the Day 1/Day 2 (pre-deployment/deployment/post-deployment) operation of DC fabric was complete when we integrated ONES with FTAS (Fabric Test Automation Suite). It is a continuous automation test suite designed for multi-vendor SONiC validation. FTAS can help with a quicker turnaround time for the pre-deployment phase. It is also integrated with ONES which gives complete visibility during the pre-deployment. You can find more information about FTAS here.

Orchestration functionality of ONES takes care of the deployment phase. Intent-based configuration YAML templates are tested and can be used for reference.

All these cutting-edge innovations only mark the initial stride towards Aviz Networks’ vision, and more is yet to come. With our strong team of support engineers, we are well-equipped to empower customers with a seamless SONiC journey using the ONES stack.

For further insights into ONES, please visit our website.

If you wish to get in touch with me, feel free to connect on LinkedIn here

FAQs

1. What is the ONES Stack and how does it simplify SONiC deployment and operations?

ONES (Open Networking Enterprise Suite) is a unified, vendor-agnostic platform that simplifies the Day 1, Day 2, and Day N operations of SONiC-based data center fabrics. It streamlines pre-deployment validation, orchestration, multi-vendor visibility, and supportability  making SONiC adoption faster, cost-effective, and easier to manage across large-scale enterprise environments.

ONES supports both SONiC and non-SONiC devices by pulling metrics using eAPIs, NVAPIs, and other interfaces, then normalizing the data for unified visibility. It enables full-fabric visualization, performance monitoring, and alerts, providing a seamless operational experience even in mixed NOS (Network Operating System) environments.

FMCLI (Fabric Manager CLI) was developed to overcome the limitations and inconsistencies in SONiC’s native CLI. FMCLI provides a flexible, consistent command-line experience to configure and manage SONiC fabrics efficiently, simplifying orchestration and supporting intent-based configurations across diverse vendor hardware.

ONES uses a dedicated telemetry layer that collects over 200+ critical metrics using gNMI and advanced monitoring beyond RedisDB. It normalizes multi-vendor telemetry data into a unified format, allowing network operators to correlate issues faster, diagnose performance gaps, and ensure proactive decision-making.

The Fabric Test Automation Suite (FTAS) integrates with ONES to automate validation of SONiC fabrics before deployment. It helps verify multi-vendor configurations, scalability, resiliency, and performance through a library of pre-validated tests, drastically reducing the risk of deployment failures and speeding up production readiness.

Categories
Open Networking Enterprise Suite SONiC

Unveiling the Backbone: Exploring ONES Network Latency Measurement Backend

Introduction

In today’s datacenter landscape, network latency holds significant importance due to its impact on the overall performance, efficiency, and reliability of datacenter operations. Here are key aspects that highlight the significance of network latency in data centers. Low network latency is crucial for ensuring optimal performance of applications hosted in data centers. Users expect fast response times when accessing services and applications, and latency directly influences the perceived responsiveness of these systems. Organizations face various challenges in measuring and optimizing network latency, as this task involves complex considerations related to infrastructure, applications, and user experience. Some common challenges include the complexity of the network infrastructure, dynamic workloads and continuous monitoring feeding to the analysis.
This blog introduces you to the backend of ONES network latency measurement component , the core engine responsible for collecting data related to network latency. This component plays a crucial role in providing insights into the performance of a network, helping organizations monitor and optimize their infrastructure. It supports various network protocols, including ICMP (Internet Control Message Protocol), and TCP (Transmission Control Protocol), depending on the need.

Let’s explore the key aspects and functionalities of the backend.

Core Features

The NWSLA measurement component provides an agent that runs in the SONiC switches as well as servers. This agent exposes an API to the ONES collector eco-system and allows for triggering the latency measurements. Latency to a Destination IP can be measured using either ICMP or TCP. The calculation involves the following parameters
  • Protocol - ICMP vs TCP
  • Destination IP
  • Port (for TCP)
The above diagram explains the same. ONES Collectors controls & facilitates the probes, allowing latency measurements to be performed by the whole ecosystem of endpoints. One of the important features of this agent is its ability to allow the calculations to be calculated periodically. For instance an operator wishes to calculate the latency between point A to point B every 5 minutes. This aids the operator in the following cases
  • Periodic latency calculations help operators monitor the health of the network. An increase in latency can indicate potential bottlenecks or issues that need optimization.
  • By measuring latency regularly, operators can verify that the network meets the latency requirements specified in SLAs, maintaining high-quality services for customers and stakeholders.
  • Regular latency measurements provide a baseline for normal network behavior. Deviations from this baseline can signal potential faults, enabling operators to troubleshoot and resolve issues promptly.
  • Latency calculations help operators understand how network performance scales with increasing loads. This information is crucial for capacity planning and ensuring the network can handle growing demands.
  • Periodic latency measurements are essential for monitoring the performance of real-time applications, such as video conferencing or voice-over IP (VoIP), where low latency is critical for a smooth user experience.

To cater to such requirements, ONES allows the operator to schedule the calculation of the latency periodically over specified time intervals. This allows the operators to understand the performance of the networks proactively.

NanoSecond Level Precision

One of the unique features of the ONES infrastructure that calculates this latency is its ability to calculate the latency in terms of nanoseconds. Calculating the latency in terms of nanoseconds offers some unique advantages, in scenarios where extremely precise and rapid measurements are essential perfect for the datacenter networks. It allows
  • Ultra-Low Latency Requirements: Nanosecond-level measurements cater to applications with ultra-low latency requirements. These applications demand the fastest possible response times for optimal performance.
  • Real-Time Communication: Applications like telecommunication services and real-time communication tools benefit from nanosecond-level measurements. This precision ensures that communication is nearly instantaneous, enhancing the quality of real-time interactions
  • Computing: In edge computing environments, where processing occurs closer to the data source, nanosecond-level latency measurements are crucial. This precision helps evaluate the effectiveness of edge deployments in providing rapid responses.
  • High Throughput Networks: In networks with exceptionally high throughput, nanosecond-level precision is advantageous for accurately assessing the efficiency and performance of the network, especially under extreme loads.
In summary, calculating latency in nanoseconds offers advantages in situations where speed, precision, and real-time responsiveness are paramount.

Decoding Nanosecond Latency Calculation

Precision in nanosecond latency calculation is a sophisticated endeavor. ONES adopts an innovative approach by modeling the network analogous to an optical channel, ensuring a high degree of precision in latency calculation. The methodology involves sending bursts of packets and deriving latency measurements seamlessly, deviating from the conventional approach of correlating request and response times. This eliminates the need to calculate latency per request before initiating subsequent probes. The flexibility of ONES allows for the configuration of parameters to align with specific network requirements.

Scalability & Robustness

The ONES ecosystem excels in delivering exceptional scalability and robustness. The philosophy of ONES is seamlessly reflected in the design of its latency measurement, ensuring scalability and robustness are prioritized. This commitment is affirmed through various validations, including measuring latency under load, the seamless addition of new nodes for calculations, the capacity to handle and sustain a significant number of probes by the agent, built-in fault tolerance features, optimized resource utilization, and consistent operational and longevity behavior.

Use-Cases: Ping-pong Mesh

One of the simple use-case scenarios will be to trigger the measurement across the endpoints of the network. This initiates the latency test from the end points attached to the network ensuring that the packets used to measure latency traverse the network to reach the other endpoint. This will be calculated proactively at a set defined interval allowing to check on the latency periodically. Identifying latency bottlenecks helps optimize resource allocation and maintain high-quality services
Under such use-cases, the measurement of latency plays a vital role in optimizing overall network performance. Low-latency communication directly enhances user experience, aids in capacity planning, facilitates proactive issue resolution, and furnishes valuable data for making informed decisions about network infrastructure.These scenarios can be expanded, including the exploration of latency between the most distant leaf nodes, and so on.

ONES Network SLA - Future Looking

Subsequent releases of ONES provide robust support with advanced features built upon this foundational base. Initially, integration with the ONES UI will be seamless, offering comprehensive cloud integration support. Additionally, support for path tracing and availability metrics will be extended across the system.

In conclusion, the backbone of a network latency measurement tool functions as the core engine responsible for gathering, processing, and evaluating data to gauge the vitality and efficiency of a network. It stands as a pivotal element for organizations aiming to sustain ideal network latency, guaranteeing a smooth and responsive user experience.

Have comments or feedback? Please feel free to get in touch with us
For experiencing SONiC, Please try ONES Center  https://aviznetworks.com/one-center
For detailed case study of SONiC, please refer here

FAQs

1. What is the ONES Network Latency Measurement Backend and how does it work?

The ONES Network Latency Measurement Backend is a core engine that enables high-precision latency monitoring across SONiC switches and servers. It uses agents deployed at endpoints to trigger periodic ICMP or TCP-based probes, providing real-time latency insights critical for monitoring network health and optimizing datacenter performance.

ONES models network latency measurement similar to an optical channel by sending bursts of packets, enabling seamless, ultra-precise latency calculations without the overhead of per-request correlation. This unique approach delivers nanosecond-level accuracy ideal for real-time communication, edge computing, and high-throughput data center networks.

Periodic latency monitoring helps detect performance bottlenecks early, ensures SLA compliance, identifies real-time communication issues, supports capacity planning, and establishes a baseline for normal network behavior — enabling proactive management and optimized user experience.

Yes, the ONES infrastructure is designed for exceptional scalability and robustness. It supports a large number of simultaneous probes, handles high-traffic loads, adds new nodes seamlessly, and maintains consistent accuracy and operational stability even in large, dynamic data center environments.

ONES can be used for real-time latency monitoring between endpoints, identifying bottlenecks across leaf-spine topologies, optimizing RoCE traffic, validating edge computing responsiveness, supporting capacity planning, and proactively detecting network issues  ensuring superior performance across modern AI and cloud-driven infrastructures.

Categories
SONiC

Be a VIZionary by Partnering with Aviz Networks for the Best in Network-Focused GenAI, Visibility, and SONiC-Focused Open Networking Solutions

The right partner is everything. The partnerships that truly last, are built on a solid foundation of shared core beliefs and principals. Principals like building and maintaining honesty and trust in your partner. Belief that partners that stick by each other, are loyal to one another, and go the “extra mile” to help each other service their customers together. Principals that govern that partners act because they do the right thing by their customers first and foremost, and for each other every step of the way.

More importantly, the right VIZionary Partner Program, built on these same foundational principals and beliefs, is the reason, our partners WILL STAY our partners.

Hardware manufacturers, System Integrators (SIs), Value Added Resellers (VARs), and distributors, have many choices when choosing the right partner to Go-To-Market (GTM) and build long-term and prosperous relationships. Aviz Networks is proud to be THAT chosen partner with A VIZionary Partner Program that embodies those same principals and beliefs in building long-term business together. A VIZionary Partner Program built on the principals for building solid partnerships, engineered to incentivize our Partners all along the way as a direct extension of Aviz, and to represent Aviz as members of our own team, not just a fulfillment mechanism.

In the last month we have proudly launched our BRAND NEW Aviz Networks VIZionary Partner Program and we are privileged to announce that we have some of the best industry partners in networking and whitebox hardware, security and networking systems integration, open-source software, and Artificial Intelligence (AI) delivery to kick off our Program.

SIs and VARs, after being embraced by Aviz Networks Executive team, culture, and solutions, are not shy about their potential for growth with Aviz.

Aviz embodies the ‘3 C’s’ they promote to their customers in everything they do: Choice, Control, and TRULY Cost Effective Solutions that bring a unique value to our current offerings, and we look forward to seeing this partnership blossom rapidly as we seek to introduce Aviz unique blend of Open Source SONiC Savvy and GenAI to the Network.

Bill Rubin, CEO and Founder, Connect IT Solutions of San Mateo, CA

We've been refining our True SONiC program to offer a vendor-neutral, end-to-end service for SONiC adoption, which includes ecosystem guidance, hardware procurement, and infrastructure design services, and we see a great synergy with Aviz Network’s extensive SONiC solutions and support capabilities. Integrating Aviz’s services into True SONiC —such as 24/7 support, custom use case testing, and your advanced Networking 3.0 solutions—into our program could significantly enhance our joint value proposition. We’ve been seeing particular interest from mid-market enterprise and startup-type customers.

Alex Cronin, CEO and Founder, Hardware Nation

Hardware partners alike have been equally impressed with Aviz attention to every detail from comprehensive SONiC-focused network Fabric Testing and Automation Service (FTAS), to joint GTM activities are likewise enthusiastic about the synergies from partnering with Aviz.

We see a promising future with Aviz as our partner. By integrating Aviz's Best-In-Class GenAI, Automation, and Visibility Solutions, alongside 24x7x365 Support for SONiC, into UfiSpace's Next-Generation complete line of SONiC-powered telecom and data center solutions, customers will have much greater choices and control over their networks, supply chain, and data all through a cost-effective combination.

Vincent Ho, CEO of UfiSpace

Come replicate the success our new partners who are already enjoying success as an integral part the Aviz Networks Team HERE (Aviz Partner Site).

Join an EXCLUSIVE eco-system of the BEST technology partners in the industry and harness the 3 Core Pillars of Aviz Solutions and Networking 3.0: Choice, Control, and Cost Savings for your customers as we help them EMBRACE the AI-fueled Networking 3.0 era because….

TOGETHER, WE NETWORK BETTER!

Categories
Fabric Test Automation Suite SONiC

FTAS 2.3 and SONiC: Elevating Data Center Interconnect Quality with EVPN and VXLAN

Fabric Test Automation Suite (FTAS) – a solid set of test cases designed with precision to assess SONiC’s readiness for deployment is ever evolving. Our ongoing dedication to getting better has shaped FTAS, growing with valuable customer input and incorporating new features in sync with the latest SONiC releases.

Exciting news on our journey – FTAS 2.3 is now live! This version comes with a bunch of improvements, all geared towards making SONiC assessment even better and smoothing out pre-deployment testing. Let’s dive into what’s new and how it can elevate your testing experience!

Let’s delve into the key enhancements introduced in FTAS 2.3

Data Center Interconnect

Data Center Interconnect (DCI) harmonizes a sophisticated integration of advanced technologies. EVPN (Ethernet VPN), VXLAN (Virtual Extensible LAN), and MCLAG (Multi-Chassis Link Aggregation Group) collectively contribute to the intricate network orchestration. EVPN dynamically manages control plane operations, VXLAN extends virtual networks seamlessly, and MCLAG ensures redundancy and load balancing. This technical synergy within DCI forms a robust infrastructure, addressing the challenges of interconnected data centers. Each technology plays a pivotal role in achieving fluidity and resilience in the overall network composition.

Based upon the customer deployments, the FTAS latest release is packed with new test cases to enhance the existing coverage.

BGP-EVPN with VXLAN

VXLAN EVPN facilitates seamless communication in data centers by extending Layer 2 connectivity over a Layer 3 network. This technology ensures efficient management of virtualized workloads, enabling dynamic learning and distribution of MAC and IP addresses. With its scalable and flexible approach, VXLAN EVPN is instrumental in modern data center deployments, providing a reliable solution for network virtualization and resource optimization.

EVPN/VXLAN Topology

Here are the VXLAN EVPN scenarios covered in FTAS 2.3

Feature Test Scenario
BGP-EVPN-VXLAN EVPN VXLAN Configuration and verification
EVPN VXLAN for known unicast, BUM traffic (eBGP) with RIF
EVPN VXLAN for known unicast, BUM traffic (eBGP) with SVI
EVPN VXLAN for known unicast traffic (eBGP) with link events and router failure - RIF
EVPN VXLAN for known unicast traffic (eBGP) with link events and router failure - SVI
EVPN VXLAN for known unicast traffic (eBGP) with link events and router failure - RPCH
Symmetric IRB with EVPN eBGP-RIF (eBGP)
Symmetric IRB with EVPN iBGP-RIF (iBGP)
Symmetric IRB with EVPN iBGP-SVI
Symmetric IRB with EVPN eBGP-RPCH
Symmetric IRB with EVPN iBGP-RPCH
Asymmetric IRB with EVPN eBGP
Asymmetric IRB with EVPN iBGP

MCLAG

MCLAG (Multi-Chassis Link Aggregation) enhances data center reliability by creating a resilient, redundant network. In the event of a switch failure, MCLAG seamlessly transfers traffic to a backup switch, preventing downtime. This technology ensures high availability, load balancing, and simplified network design, making it indispensable for critical applications where continuous connectivity is paramount.

L2 MCLAG
L3 MCLAG

FTAS 2.3 covers following MCLAG use cases based upon the customer deployments.

Deployment Test Scenario
L2 MCLAG MCLAG-L2 Steadystate with portchannel and vlan on peer link
MCLAG-L2 Member Link Down
MCLAG-L2 Peer Link Down
MCLAG-L2 Active Reboot
MCLAG-L2 Standby Reboot
L3 MCLAG MC-LAG L3 validation using portchannel configuration
MC-LAG L3 validation, Bring down the member link of Spine1
MC-LAG L3 keepalive link down
MCLAG-L3 Active Reboot
MCLAG-L3 Standby Reboot

ECMP Scaling & Resilience

ECMP (Equal-Cost Multipath) optimizes network performance by distributing traffic across multiple paths of equal cost. In data center scenarios, ECMP enhances load balancing, increases bandwidth utilization, and ensures fault tolerance. This technology is particularly valuable in modern, dynamic environments where efficient resource utilization and high availability are crucial for delivering reliable and responsive services.

The support for ECMP scalability was added in FTAS 2.2. Based upon further customer use cases, FTAS 2.3 adds the support for eBGP to ECMP scalability. 

Here is the list of ECMP eBGP scalability scenarios added in FTAS 2.3

Test Scenario Description
ECMP Scalability - eBGP support Scalability to 16 ECMP paths and 32K eBGP routes
Scalability to 32 ECMP paths and 64K eBGP routes
Scalability to 64 ECMP paths and 128K eBGP routes
Scalability to &lt number of leaf1 ports > ECMP paths and 1K eBGP routes per path

Summary

FTAS 2.3 is here, packed with enhancements for seamless SONiC assessment. The latest release introduces new VXLAN EVPN scenarios, covering various test cases for efficient data center communication. Additionally, MCLAG use cases ensure resilient, redundant networks for critical applications, preventing downtime during switch failures. 

FTAS 2.3 also extends support for ECMP scalability with eBGP, enhancing network performance and fault tolerance. Explore the specifics of these upgrades, from VXLAN EVPN configurations to MCLAG and ECMP eBGP scalability scenarios. Elevate your testing experience with FTAS 2.3’s comprehensive coverage and improved capabilities.

Schedule a Call with Our Team to Delve into FTAS. For comprehensive information before the scheduled call, visit our FTAS product page.

FAQs

1. What is the role of FTAS 2.3 in validating SONiC-based Data Center Interconnect (DCI) architectures?

FTAS 2.3 offers a comprehensive set of test cases covering EVPN, VXLAN, MCLAG, and ECMP scenarios, ensuring that SONiC deployments are fully validated for data center interconnect use cases. It helps assess network scalability, redundancy, and resilience in real-world conditions before production rollout.

FTAS 2.3 introduces extensive test cases for VXLAN EVPN scenarios, enabling thorough validation of Layer 2 over Layer 3 connectivity, MAC/IP address learning, link failure recovery, and router failure handling — essential for modern, virtualized data center fabrics.

FTAS 2.3 covers L2 MCLAG and L3 MCLAG test cases, including steady-state validation, member and peer link failures, active/standby reboots, and keepalive link monitoring, ensuring high availability and fault tolerance in multi-chassis link aggregation setups.

 FTAS 2.3 expands ECMP testing by adding eBGP-based scalability scenarios, validating support for up to 128K routes and up to 64 ECMP paths, ensuring that SONiC can efficiently handle dynamic, high-bandwidth, load-balanced data center traffic environments.

FTAS 2.3 provides enterprises with a robust, customer-driven testing framework to validate SONiC readiness, optimize network resilience, guarantee scalability, and minimize deployment risks  accelerating time-to-production and boosting overall confidence in SONiC adoption

Yes, FTAS 2.3 includes specific test cases that simulate link events, router failures, and peer link down conditions across VXLAN EVPN and MCLAG topologies, helping enterprises validate SONiC’s resilience and fast recovery under real-world failure conditions.

FTAS 2.3 enables testing across hardware platforms from different vendors running SONiC, verifying protocol consistency, scalability, and failover behavior. This ensures enterprises can confidently build multi-vendor, open networking fabrics without facing hidden interoperability challenges.

Absolutely. With support for 64-way ECMP, 128K routes, and robust EVPN/VXLAN and MCLAG validations, FTAS 2.3 is well-suited for high-density AI/ML workloads that require massive bandwidth, low latency, and resilient connectivity between compute nodes and storage fabrics.

FTAS 2.3 supports testing common data center topologies such as Leaf-Spine, Clos networks, EVPN-VXLAN overlays, L2 and L3 MCLAG setups, and ECMP-scaled fabrics — covering both traditional and modern cloud-scale architectures.

By systematically validating routing convergence, failover handling, load balancing, and scalability before production rollout, FTAS 2.3 identifies potential configuration or performance gaps early, reducing the chances of unexpected network outages after deployment.

 

Categories
SONiC

AI Networking Fabric with SONiC: The Future

AI Networking Fabric with SONiC The Future

In recent times, the conversation around AI infrastructure has intensified, with leading companies globally gearing up to enhance their AI clusters in anticipation of the skyrocketing AI traffic in the years to come. Amidst this surge, the dialogue often turns towards the efficiency and cost-effectiveness of AI workloads, challenging the traditional reliance on public clouds due to total cost of ownership (TCO) considerations.

Why is SONiC becoming the preferred choice for AI networking fabric?

As a leader at Aviz, I’ve gained insights into the strategies of switching vendors and how customers respond to them. Generally, customers aim to achieve the following high-level goals, and SONiC aligns perfectly with these objectives:

  • Open Choices, Control, and Cost Savings: SONiC brings the 3Cs: Choices, Control, and Cost Savings. It breaks free from vendor lock-in, empowers network control (akin to Linux for computing), and offers significant savings for all other Opex and CapEx investments – including GPU investments or app development redirection.
  • Accelerated Deployment: In the fast-paced AI landscape, switch ASIC vendors race to develop chips optimized for AI traffic. SONiC, more often is the first NOS which gets ported onto these new ASICs due to its wide usage by hyperscalers, this priority of SONiC porting ensures swift access to the latest ASIC technology along with vendor-independently.
  • Network Innovation by using the data: Concerns about GPU cycle cost-effectiveness and network bottlenecks are common. However, SONiC’s open-source, microservices architecture makes it easier to customize your network to synergize with GPU cycles, fostering network-based innovations. The biggest advantage is to easily get the normalized data set where-in you can use the AI to manage your AI Networking fabric. That is only possible if you have a vendor agnostic normalized data. SONiC provides you that foundation where you can use AI to manage your AI fabric in future.

Aviz, in collaboration with industry giants like NVIDIA, Cisco, Edgecore, Wistron, and Celestica, stands at the forefront of AI fabric innovations. SONiC (Software for Open Networking in the Cloud) is emerging as the optimal choice for AI networking fabric due to its deep integration with leading partners and transformative capabilities.

3 Reasons Why Now is the Perfect Time for SONiC-Based AI Networking

Customers who have successfully deployed SONiC for their AI fabrics share several common reasons for choosing SONiC, including conducting rapid PoCs and swiftly transitioning to production. The urgency for adopting SONiC in AI infrastructure is highlighted by several converging trends:

  • Alignment with GenAI Revolution: The SONiC revolution dovetails perfectly with the rapid advancements in Generative AI (GenAI), offering a robust and scalable networking foundation that can adapt to the evolving requirements of AI workloads.
  • Readiness for Multi-Tenancy: With the necessary protocols for supporting multi-tenancy already in place, SONiC is poised to accommodate the complex networking demands of today’s AI-driven environments, ensuring efficient resource allocation and isolation.
  • Future-proof Networking for you AI Workloads: With support for speeds up to 800GbE, SONiC is at the forefront of networking technology, designed to handle the intensive data throughput required by AI applications. Industry leaders like Cisco, Broadcom, Marvell and NVIDIA are championing SONiC’s capabilities, highlighting its suitability for the most demanding AI tasks.

SONiC Solution Selection: 3 Key Steps for Making the Right Choice

Open-source-based solutions can sometimes be confusing, especially when most hardware vendors offer their own legacy solutions alongside open-source options, with an obvious inclination towards locking you into their legacy systems. Selecting the appropriate SONiC solution requires a careful assessment of your needs and the capabilities of the available offerings:

  • Opt for Open Source and Proven Solutions: An open-source, well-tested SONiC solution guarantees transparency and flexibility. Closed-source options might not offer the full benefits of SONiC, relegating them to the same status as proprietary systems.
  • Demand Comprehensive End-to-End Testing: Ensure the solution has undergone rigorous end-to-end testing, covering your specific traffic types and simulated scenarios. This step is crucial for verifying the solution’s reliability and performance in real-world conditions.
  • Seek Multi-Vendor Compatibility: A true SONiC solution should support a broad ecosystem, offering seamless integration and compatibility with hardware from various vendors. This flexibility ensures that your networking infrastructure can evolve alongside your AI initiatives without being hindered by vendor lock-in.

Starting by Leveraging SONiC Ecosystem with Aviz Networks

Another challenge with open source is determining how and where to test the end-to-end solution and evaluate each option using a consistent standard. This process should be straightforward, cost-free, and quick. Starting your SONiC journey with a community lab created by all leading vendors and Aviz Networks means becoming part of an ecosystem designed for inclusivity and extensive support:

  • Embrace a Comprehensive Ecosystem: Aviz Networks champions a holistic approach to SONiC adoption, bringing together diverse hardware vendors to create a unified and robust AI networking fabric. This ecosystem is designed to support the varied and dynamic needs of AI applications.
  • Leverage the ONE Center for Evaluation: Aviz ONE Center serves as a hub for customers to evaluate various SONiC solutions in a real-world, multi-vendor environment. This facility allows for hands-on testing and comparison, ensuring that you can make informed decisions without any initial investment.
  • Seamless Proof of Concept: Start your SONiC deployment journey with a straightforward Proof of Concept (PoC) phase at our ONE Center. This phase allows you to experience the capabilities of SONiC firsthand, ensuring that the solution meets your specific requirements before committing to a full-scale deployment.

What’s Next?

Interested in diving deeper? Join a conversation with our customers and discover how Aviz Networks has facilitated successful RoCE-based deployments leveraging SONiC. Explore the various options available and understand the cost implications. In these transformative times, Aviz Networks, alongside our vibrant partner ecosystem, is dedicated to mainstreaming SONiC as the networking OS of choice, mirroring Linux’s success in the operating systems domain.

Seize the opportunity to shape the future of networking. Contact us now to start the conversation and embark on your journey towards unparalleled networking performance and efficiency!

FAQs

1. Why is SONiC considered the best choice for building AI networking fabrics?

SONiC offers open-source flexibility, vendor neutrality, and deep hardware integration, enabling rapid adoption of next-gen ASICs and support for ultra-high speeds like 800GbE. Its microservices-based architecture, multi-vendor ecosystem, and real-time data normalization make it ideal for managing intensive AI workloads at scale.

By eliminating vendor lock-in, enabling competitive hardware selection, and allowing customizable network control, SONiC significantly lowers both CapEx and OpEx. Enterprises can redirect savings towards GPU investments, AI application development, and scaling infrastructure without being constrained by proprietary networking costs.

Enterprises should prioritize proven open-source SONiC solutions, insist on comprehensive end-to-end traffic testing, and ensure multi-vendor compatibility. This strategy guarantees flexibility, transparency, and long-term scalability for rapidly growing AI networking needs.

Enterprises can use facilities like the Aviz ONE Center to conduct real-world, multi-vendor Proof of Concept (PoC) testing. This hands-on evaluation ensures the SONiC solution meets performance, compatibility, and scalability requirements  all without any initial hardware purchase or financial commitment.

SONiC’s modular, open-source architecture allows rapid innovation, AI-driven network optimization, and seamless adaptation to newer ASICs and protocols. Its proven ability to handle multi-tenancy, massive scalability, and vendor-agnostic deployments ensures readiness for the next-generation GenAI revolution.

SONiC supports advanced features like Priority Flow Control (PFC), ECN (Explicit Congestion Notification), and RoCE-specific optimizations, ensuring low-latency, high-throughput performance critical for AI/ML training clusters and data center interconnects handling GPU-based workloads.

SONiC supports cutting-edge network speeds up to 800GbE, seamlessly integrating with modern switch ASICs from vendors like Broadcom, NVIDIA, Marvell, and Cisco. This enables enterprises to build ultra-high bandwidth, scalable AI fabrics ready for next-gen data demands.

Yes, SONiC fully supports multi-tenancy using protocols like VRF, VXLAN, and EVPN. This allows organizations to isolate workloads, efficiently allocate network resources, and ensure secure segmentation across AI tenants and dynamic cloud-based AI applications.

Major industry players such as Microsoft Azure, Alibaba, Tencent, and NVIDIA are actively using SONiC in their AI and hyperscale environments. Increasingly, enterprises and service providers are following suit to achieve flexible, scalable, and cost-effective AI networking solutions.

SONiC’s microservices architecture and open telemetry capabilities create normalized datasets across the network. This foundation enables integration with AI/ML models for real-time anomaly detection, predictive maintenance, capacity optimization, and autonomous AI-driven network operations.

Categories
SONiC

Debunking 5 Common Myths About SONiC

In the ever-evolving world of networking, SONiC NOS (Software for Open Networking in the Cloud) has emerged as a game-changer for data center and edge networks. However, like any technology, SONiC has not been immune to myths and misconceptions. When we talk to users evaluating SONiC for their network, we often hear… “SONiC is not ready for us, at least just yet”. In this blog, we’ll debunk five common myths about SONiC.

1. SONiC is not ready for the Enterprise

Myth:  One of the most common things we hear is that SONiC does not have features such as EVPN, or MC-LAG, and users are not comfortable about its quality standards. It is also difficult to integrate with our existing NetOps tools.

Reality: The table below gives a good high-level overview of what is already available for use cases across various network architectures. This list is evolving very fast, and if you follow the SONiC community then with-in next few months this list will increase in use cases, as well as in vendors who support those use cases.

Use Case Essential Protocols Supported By
IP CLOS  BGP Unnumbered Most vendors
AI Fabric RoCE Several vendors
IP CLOS w/EVPN EVPN T1 to T6 Several vendors
Edge & Campus MLAG, 802.1x, STP, POE Several vendors

To see evidence of whether or not SONiC has the support for the above capabilities, and try them out on any hardware sourced from ANY of the white-box vendors, you can schedule a demo in our ONE Center. We will not only walk you through each of the available options, but also show how SONiC can be easily integrated within your existing NetOps environment. 

At Aviz Networks we help accelerate SONiC Community feature development, and we work with customers to develop the features they need and give them back to the community. Essentially, Aviz is bringing a paradigm-shifting new concept to market called “Roadmap By-pass”, where customers get the features they need, when they need it,  in a more agile fashion instead of a single vendor prioritizing their own desires or for the “highest bidder”.

2. Deploying Community SONiC requires an army of Network Engineers

Myth:  One of the most persistent myths about SONiC is that it is exclusively designed for hyperscalers who have access to a large pool of network engineering expertise in-house. While the NOS offers powerful networking capabilities, it’s not a simple-to-deploy solution for all environments.

Reality: Yes, SONiC does require a good understanding of networking principles, SDN concepts, and the knowledge of hardware compatibility, but the fact is a vast majority of large to medium enterprises have already deployed SONiC using their existing teams with an added bonus of a little external help. Companies like eBay, have benefited immensely both in terms of cost savings and the flexibility of innovation by deploying SONiC on white-box switches. The market leadership brief by Futuriom lists down how Networking users can accomplish similar feat. According to publicly accessible information on the SONiC community page, numerous enterprises including Comcast, eBay, Target, and Bloomberg have either deployed SONiC or are currently in the process of doing so.

At Aviz, we have been helping enterprises of all sizes with purpose-built automation tools and data normalization solutions that make the SONiC deployment and operations quick and smooth. 

You can discover the magic behind eBay and some of our other rapidly expanding SONiC deployments by checking out our ‘Open Networking Enterprise Suite (ONES)’ and requesting a demo. 

ONES is not only helping large enterprises with their SONiC deployments, but dozens of small to medium enterprises in their own migration to SONiC. So, in reality, you don’t  need an army of in-house engineers with PHDs, so long as you leverage the right resources and partners like Aviz who is dedicated to your success and SONiC’s future.

Recently we went one step further to put AI to use for easier migration towards SONiC. Our latest solution Network Copilot™ is designed to generate templates for easy transition.

3. Community SONiC POCs are challenging, and procurement process is a mess

Myth:  A lot of people assume that conducting Proof of Concepts (POCs) with SONiC can be a challenging endeavor, often compounded by a complex and cumbersome procurement process. The versatility SONiC offers for a diverse array of supported hardware and configurations, make it challenging to select the right components for your use cases. Additionally, navigating the procurement process, which involves evaluating multiple vendors, and negotiating contracts, can be overwhelming and time-consuming.

Reality: SONiC’s open-source nature allows organizations to break free from vendor lock-in and choose best-of-breed components from various hardware vendors, tailoring their network to specific needs. Conducting a multi-vendor POC with SONiC can appear to be a complex undertaking, especially when you first start to think about what it takes for a successful deployment:

    • Selecting hardware vendors

    • Testing for performance and feature requirements

    • Scalability per your specific use cases

    • Ensuring interoperability with your existing ecosystem of infrastructure and tools

Thankfully, you don’t need to reinvent the wheel or spend time and money doing it. Aviz has done all the heavy lifting for you by partnering with nearly all major hardware vendors, to create an environment where any organization can conduct POCs across a wide range of options without having to invest a single dollar in purchasing hardware for POCs. 

The Aviz ONE Center is a perfect solution for low-effort, low-cost SONiC POCs with online and in-person access to try out SONiC capabilities on any hardware of your choice. Also, Aviz has created tools (also available in ONE Center) that can perform hundreds of tests for performance and scalability pertinent to your specific use cases. 

Additionally, today, many System Integrators (SIs) provide end-to-end solutions for SONiC, which means, you do not need to buy hardware, software, and support separately from multiple vendors.

In fact, many of our own SI partners have simplified the procurement process so much that they can generate a single BOM (bill of materials) across 10+ vendors and ensure that your procurement teams don’t have to interface with them individually.

4. Community SONiC is not secure because it’s open-source

Myth:  Many claim that SONiC is inherently insecure simply because it is open-source and exposes the source code to potential attackers. 

Reality: Exactly the opposite. Practically every Security Professional Aviz has encountered lobbies for Open Source because of the inherent review the software is given, and thus Open Source itself does not equate to insecurity. The security of open-source software depends on various factors, just like any other software implementation. While Open Source software by nature exposes the source code to everyone, which includes potential attackers, it also invites a large pool of experts to discover and address security vulnerabilities. The transparency of open-source projects often means vulnerabilities are identified and fixed faster than in proprietary software. 

The security of any software deployment or implementation, including SONiC, primarily depends on how well it is configured, maintained, and monitored. A vast community of SONiC developers actively scrutinizes and contributes to the codebase, ensuring that any security concerns are swiftly addressed, making SONiC more resilient to vulnerabilities, and substantially reduces the time to review and approve its use for a wide range of applications and services.

5. Open Source SONiC is just a short-term trend

Myth:  Some suggest that SONiC is merely a short-term trend and will fade away in the near future. Several open-source and commercial Open NOS solutions existed before SONiC, gained prominence, but eventually got consumed into proprietary networking ecosystems.

Reality: The consistent growth in the adoption of SONiC across large, medium, and small enterprises, combined with support and stewardship by major players like The Linux Foundation (LF), Open Compute Project (OCP), Microsoft, and the vast majority of networking hardware vendors, is a testament to its staying power. 

SONiC’s modular and open-source approach to networking addresses many of the pain points organizations have been facing for decades, making it not just a trend, but a significant shift in the networking paradigm. As organizations seek more control over their network infrastructure and adapt to evolving cloud-based architectures, SONiC offers a robust solution to both challenges.

Additionally, SONiC has already garnered unparalleled support from the vast majority of networking vendors. This is the first time in history, that an Open-source NOS has been whole-heartedly embraced by networking giants like Arista, Broadcom, Cisco, Juniper, Marvell, NVIDIA, along with numerous white-boxvendors such as Celestica, Edgecore, Quanta, Super Micro, and growing every day.

In their recent 2023 report Gartner explicitly stated “Open networking has been replaced on the Hype Cycle with SONiC, which garners the most client interest of any open networking technology.” 

A 5 Point Analysis on why Gartner Preferred SONiC Over Open Networking in their 2023 Hype Cycle is a great read that clearly  explains this ongoing trend. 

While technology trends may come and go, SONiC’s foundations align with the enduring demand for adaptable, cost-effective, and scalable networking solutions, suggesting that this is more than just a short-term trend in the networking landscape.

Conclusion

SONiC’s ascent in the networking sphere is undoubtedly well-deserved, poised to transform organizational networking strategies. By dispelling prevalent myths, we aim to offer clarity and inspire more entities to explore the myriad benefits SONiC offers. While due diligence is essential in adopting any technology, SONiC presents compelling options for next-generation networks. Understanding its true potential empowers organizations to make informed decisions about integrating SONiC into their network infrastructures.

Reach out to reserve your spot at our ONE Center for a proof of concept, where various vendor solutions, including Cisco SONiC, NVIDIA SONiC, Celestica SONiC, Marvell SONiC, Wistron SONiC, Edgecore Community SONiC, Supermicro SONiC, Enterprise SONiC etc, can be thoroughly explored and tested, ushering in a new era of networking innovation. As a common knowledge, these vendors are helping the community innovators such as Microsoft for Azure SONiC deployments as well hence bringing the same SONiC for other deployments.

FAQs

1. Is SONiC ready for enterprise networks and complex data center deployments?

Yes, SONiC is enterprise-ready and supports essential protocols like EVPN, MC-LAG, BGP Unnumbered, RoCE, and more. It is already deployed by major organizations like eBay, Comcast, and Bloomberg, and is rapidly evolving with contributions from a large community of vendors and developers.

No, with the right tools, support, and automation solutions like Aviz ONES, enterprises of all sizes  even with existing teams can successfully deploy and manage SONiC. Many organizations have streamlined their SONiC migration without requiring massive engineering teams.

 No, SONiC’s open-source model actually enhances security. With a global community constantly reviewing the code, vulnerabilities are often identified and patched faster than in closed-source systems. Proper configuration, monitoring, and maintenance further strengthen SONiC’s security posture.

Not anymore. Through Aviz’s ONE Center and its partnerships with major hardware vendors, enterprises can easily conduct SONiC POCs across multi-vendor platforms without upfront hardware purchases, simplifying the evaluation and procurement process significantly.

SONiC is a major long-term shift. Backed by Microsoft, OCP, and major networking vendors like Cisco, Broadcom, and Arista, SONiC addresses critical needs in flexibility, scalability, and cost-effectiveness — making it the future foundation for open, cloud-first networking strategies.

SONiC offers greater flexibility, vendor neutrality, faster innovation through open-source contributions, and lower operational costs compared to traditional proprietary NOS solutions. It empowers organizations to avoid vendor lock-in and adapt their networks to evolving cloud-native and hybrid architectures.

SONiC is ideal for IP Clos data center fabrics, AI/ML training networks with RoCE, campus and edge networks using MLAG and 802.1x, and cloud-scale networking with EVPN architectures. It supports a wide range of mission-critical enterprise networking needs across various industries.

Yes, SONiC is built on a modular and standards-based architecture that allows easy integration with existing NetOps frameworks, observability tools, SDN controllers, and automation platforms — especially when using solutions like Aviz ONES that normalize data across diverse environments.

Aviz Networks provides purpose-built automation, multi-vendor support tools like ONES, simplified POC environments through the ONE Center, AI-driven migration assistance via Network Copilot, and deep engineering expertise to make SONiC deployments faster, smoother, and more efficient.

Enterprises should assess hardware compatibility, select appropriate SONiC builds (community or vendor-optimized), ensure staff readiness for Layer 2/3 operations, validate integration with NetOps tools, and engage experienced partners like Aviz to support seamless deployment and Day 2 operations.

Contact Us

Debunking 5 Common Myths About SONiC

In the ever-evolving world of networking, SONiC NOS (Software for Open Networking in the Cloud) has emerged as a game-changer for data center and edge networks. However, like any technology, SONiC has not been immune to myths and misconceptions. When we talk to users evaluating SONiC for their network, we often hear… “SONiC is not […]