Aviz Networks

10/16/2026

Multi-tenant AI networking that never disagrees with itself.

Create a tenant once, and the same network identity follows it end to end. NVIDIA Infra Controller (NICo) and Aviz ONES keep multi-tenant GPU infrastructure on one consistent tenant network model.

- NICo owns the identity: it configures the BlueField-3 DPU with the tenant’s VLAN, VNI, and VTEP information.
- ONES consumes it: it connects to NICo as a provider and retrieves those values instead of creating its own.
- ONES extends it: it applies them to the border leafs, Aviz Service Gateway, firewall, and storage-facing network.
- The result: the same tenant identifiers from the GPU host through the fabric, with a single source of truth.

Read the blog: https://na2.hubs.ly/H08rw7z0

10/15/2026

Aviz is at OCP, and one question keeps coming back for AI clusters: which packets did your network drop?

Large GPU clusters run on network connectivity, and interface counters only register that packets were discarded. sFlow Mirror-On-Drop (MOD) in SONiC 202611 is built to show what gets dropped.
- A counter confirms that drops occurred
- Engineers still need to know which packets, where, and why
- Pinning down the packet and the point of failure is what turns packet loss into a fixable problem

Read the blog: https://na2.hubs.ly/H08ppKQ0

10/14/2026

One week to go until NVIDIA GTC Berlin. A question for GPU operators: how much more could your GPUs earn with the right network behind them?

Aviz Networks will be at its first European GTC with a direct answer. Automate the network behind your GPUs, and the same infrastructure earns more.
- Faster tenant onboarding, so revenue starts flowing sooner
- Automation takes manual network work off your team’s plate, so margins get stronger on every GPU

Meet us October 20 to 22 at STATION-Berlin and Tempodrom.

Lock in time with our team before the week runs out: https://na2.hubs.ly/H08nqJ70

10/13/2026

VXLAN EVPN is the standard enterprises use to build data center and campus networks. is finally a viable option for running it.

Aviz Certified Community SONiC (CCS) release 1.2 brings VXLAN EVPN at enterprise scale to Wistron switches with Marvell Teralynx ASICs.

- CCS 1.2 supports up to 4092 L2 VNIs and 64 L3 VNIs, so you can plan for growth in tenants, segments, and endpoints from day one.
- Marvell’s latest Teralynx SAI adds VXLAN EVPN features and scale, Wistron brings proven switching platforms, and Aviz builds, validates, and supports the CCS images as your single support provider for the whole stack.
- CCS uses standards-based BGP EVPN and interoperates with existing EVPN infrastructure, so you can migrate in phases instead of forcing a risky cutover.

Read the blog below-https://na2.hubs.ly/H08l-bP0

10/13/2026

"We were looking for partners, and Aviz was exactly the right company. Aviz was trying to solve the same problem we were trying to solve, and we've done deployments together from day one." - Prathyaya Bhandarkar, Senior Director of Product Management, Marvell Technology.

Prathyaya joined us on the latest episode of our podcast series to talk about the Aviz-Marvell partnership and what it takes to turn silicon into a production-ready AI network.

A few things covered across the full episode:
- Marvell provides high-performance switch silicon. Aviz takes it from day-minus-one preparation through day-0 deployment to full production readiness, so Marvell can stay focused on silicon.
- That partnership isn't just a launch-day handoff. Aviz has run these deployments with Marvell since day one.
- Our AI-NOC brings AI operations on top of the existing infrastructure, catching a rising issue, like queue sizes climbing on a switch, before it turns into an outage, and turning idle GPU time back into revenue.
- Even after go-live, it's a real team behind it, not a support queue, which is what "production-ready" actually means in practice.

Watch the full podcast: https://na2.hubs.ly/H08lpmw0

10/12/2026

Aviz and Marvell are partnering to bring ready-to-deploy, Certified Community SONiC to AI Factory networks! Paired with Aviz ONES and its 24×7 SLA-backed support, the solution delivers end-to-end GPU-first network automation, lifecycle management, token visibility, inference KPIs, and agentic AI-powered operations across Marvell’s Ethernet switching portfolio. Read more below- https://na2.hubs.ly/H08kbmm0
https://na2.hubs.ly/H08kbmq0

10/08/2026

One week to go for our joint bootcamp with NVIDIA. Validate your AI Factory architecture on NVIDIA DSX Air before you make any spend!

Session 3 of the "Build the AI Factory Before You Build It" bootcamp goes live on October 13 at 10 AM PT. NVIDIA leads with a close look at DSX Air and AI Factory simulation, then Aviz shows how to put it to work validating your own architecture.

What NVIDIA will cover:
- An introduction to NVIDIA DSX Air
- How NVIDIA AI Factory Simulation works
- The key features of DSX Air
- The resources available to get started with NVIDIA DSX Air

What Aviz will cover:
- How the ONES Day 0-2 design launches NVIDIA DSX Air to validate your architecture before deployment
- Day-0 through Day-2 operational workflows
- The two steps from validated design to live deployment

DSX Air is normally a paid platform, and Aviz covers it for your validation window.

Register for the Bootcamp: https://na2.hubs.ly/H08g_p20

10/07/2026

The countdown to NVIDIA GTC Berlin is on. So is the question every GPU operator should be asking: how much revenue is your network holding back?

Aviz Networks is heading to Berlin for our first European GTC with a clear case to make. Automate the network behind your GPUs, and the same infrastructure earns more.

- Faster tenant onboarding means revenue starts flowing sooner
- Less manual network work for your team means stronger margins on every GPU

Meet us October 20 to 22 at STATION-Berlin and Tempodrom. Lock in time with our team before the countdown runs out: https://na2.hubs.ly/H08fWwn0

10/06/2026

For teams operating sovereign AI clouds, control has to extend into everyday operations.

That means provisioning infrastructure, keeping tenants isolated, understanding usage and troubleshooting across the AI stack.

Our latest blog explores three practical priorities:
• Provisioning and tenant isolation: how Aviz ONES and Service Gateway support infrastructure onboarding, tenant isolation through APIs and service-edge networking.
• Usage visibility: how Aviz AI-Parser identifies token usage by tenant, application or API key, alongside time-to-first-token.
• Troubleshooting: how Aviz AI-NOC, available in early access, connects network, GPU, workload and inference telemetry, with on-premises private-LLM deployment for air-gapped environments.

Read the blog below- https://na2.hubs.ly/H08ds330

For those operating sovereign AI clouds, which takes the most hands-on work today: tenant provisioning, usage visibility or troubleshooting?

Latest insights and technical articles from Aviz Networks

10/02/2026

三井情報、米Aviz Networksとリセラー契約を締結
AI時代の事業拡大を支えるオープンネットワーク運用基盤を支援

Aviz and MKI partner to drive network transformation in Japan across Al factories, data centers, and telecommunication networks.

Vishal Shukla, Co-Founder and CEO, Aviz Networks: "From the construction of AI Factories, to the strengthening of network bandwidth, to operational efficiency improvements leveraging AI, we are now in an era that will determine infrastructure transformation. Aviz is driving this transformation in AI Factory networks, data center networks, and telecommunications networks. MKI shares this vision, and the company's network engineering expertise and deep understanding of the Japanese market make it an important partner in delivering these capabilities in Japan."

Read the full release: https://na2.hubs.ly/H088ST70
Read the blog: https://na2.hubs.ly/H088ST80