Arista: AI and cloud networking portend bright future, but high-level execution required

Arista Networks is on an historic roll. Under the leadership of CEO Jayshree Ullal, the company reached $5.6 billion in annual revenue in 2023 (up a staggering 42%) on the strength of its core business: selling high-performance switching hardware and networking software to hyperscalers, social media companies and large data center customers.

Arista has now set its sights on hitting the $10 billion mark with an ambitious strategy that will expand its product line to include routers, extend its markets to campus, WAN and cloud networking, and lean heavily into AI and machine learning.

The company has carved out a unique niche in the highly competitive networking market that is dominated by powerhouses like Cisco and Juniper. Those vendors offer soup-to-nuts portfolios of networking and security products and services. Arista has built its reputation on the speed and performance of its switches, as well as its software – specifically its Linux-based Extensible Operating System (EOS) and CloudVision management system – that enable customers to build and manage flexible, scalable, heterogeneous, software-defined networks.

“They position themselves for specific missions and verticals (banking and finance) rather than for the broad market, and that lets them focus on issues that are highly relevant without looking like they’re trying to build ‘god-boxes,’” said Tom Nolle, founder and principal analyst at Andover Intel.

Arista doesn’t pretend to make all the products an enterprise customer needs, Nolle said, which creates a bit of a conundrum. “That means they need to let competitors in, and those competitors make more of the infrastructure most users need than Arista does.” Nolle suggests that Arista should try to “expand their total addressable market.”

The company is well on its way to doing just that, with several major announcements in 2023 that add routing and security features to the product lineup and target growth areas like cloud networking, campus switching and generative AI.

Arista builds on core strengths

“We are a networking-first company founded on software principles of quality and reliability. That hasn’t changed over the years,” said Martin Hull, Arista’s vice president, cloud, platform product management and systems engineering.

A core strength for Arista is EOS, which was a radical departure for the industry when it was introduced. EOS utilized software programming – the beginnings of software-defined networks – to boost the customization, performance and control of network gear.

Today, EOS manages Arista switches and routers in all environments. “Cloud architectures built with Arista EOS scale to hundreds of thousands of compute and storage nodes, with management and provisioning capabilities that work at scale,” according to the company. “Through its programmability, EOS enables a set of software applications that deliver workflow automation, high availability, unprecedented network visibility and analytics, and rapid integration with a wide range of third-party applications for virtualization, management, automation and orchestration services.”

The ability to build and utilize a single operating system from the data center, across the campus, to the cloud is very attractive to enterprise users, and will continue to drive Arista in the future, experts say. On top of that, EOS can be deployed on white-box networking hardware, which gives customers options to avoid vendor lock-in.

Arista also supports open-source programming, and its boxes can run with the Linux-based Software for Open Networking in the Cloud (SONiC), which decouples network software from underlying hardware and can run on hundreds of switches and ASICs from multiple vendors.

“EOS is one of Arista’s greatest strengths because it offers a single image with flexible packaging options – unlike its major competitors who have multiple networking operating systems,” said Sameh Boujelbene, vice president of the Dell’Oro Group.

Another core strength for Arista is CloudVision, its centralized cloud-based management platform. CloudVision provides wired and wireless visibility, orchestration, provisioning, telemetry, automation and analytics across the data center, campus, and IoT devices on edge networks.

CloudVision provides a comprehensive view of the entire network, including all devices, connections, and traffic flows. This lets network analysts quickly identify and troubleshoot problems, as well as monitor performance and resource utilization, according to Arista. In addition, CloudVision’s network information can be shared by Arista networking partners such as VMware, Microsoft and others offering management options for customers.

The idea of a single package that can manage everything across the enterprise makes it a valuable tool for customers who use multiple products to handle management of multifaceted environments, Boujelbene said. “This also makes it well-suited for deploying and managing large, complex cloud networks,” Boujelbene said.

The cloud networking push

Arista is making a concerted effort to leverage EOS and CloudVision to support cloud networking, which offers connectivity and control between applications and workloads across clouds, cloud services, on-premises data centers, and edge networks.

“We believe that cloud computing represents a fundamental shift from traditional legacy network architectures,” Arista stated. “As organizations of all sizes have moved workloads to the cloud, spending on cloud and next-generation data centers has increased rapidly, while traditional legacy IT spending has grown more slowly. Our cloud networking platforms are well positioned to address the growing cloud networking market, and to address increasing performance requirements driven by the growing number of connected devices, as well as the need for constant connectivity and access to data and applications.”

The market for cloud networking solutions is highly competitive, but analysts give Arista high marks against its major rivals – Cisco, Juniper, and VMware.

“Arista offers a comprehensive cloud networking solution that differentiates itself from others with its ML-based troubleshooting capabilities,” according to a recent cloud networking report from GigaOm. “It also has a strong focus on network optimization and operations.”

According to GigaOm, CloudVision provides a consistent operational model across domains to simplify network operations with a single orchestration tool. Key features include the ability to enable multicloud path optimization using dynamic path selection, prioritizing production traffic over non-critical traffic, and control over networking policies.

“To achieve this optimization, CloudEOS instances auto-discover the available paths and automatically establish IPsec-based data plane encryption. For optimized forwarding and dynamic path selection (DPS), CloudEOS measures delay, latency, loss, and bandwidth for each potential path, and then applies this data in real-time to determine which path to use,” GigaOm stated.

Another key feature is that CloudEOS integrates with tools such as Terraform, Ansible, Puppet, and Chef, enabling users to declaratively provision and configure public cloud environments, GigaOm wrote.

The cloud networking market is expected to grow rapidly in the coming years. The market for cloud-based IaaS networking will reach $19.4 billion globally in 2023, according to IDC, with a compound annual growth rate of 28% projected through 2026. Increasing cloud-native application architectures, distributed workloads, and their respective integration needs are the biggest drivers of IaaS cloud networking adoption, IDC stated.

Arista adds routing capabilities

Arista does face a number of challenges, including serious competition from larger and more established players, such as Cisco and Juniper. Beyond that, Arista has been very focused on its highly successful high-end markets, in particular service providers and hyperscalers that account for more than 50% of the vendor’s revenue.

But it doesn’t have a reputation for a wide breadth of products in areas such as security, SD-WAN and SASE compared to its competitors. However, that is changing. Arista now supports SD-WAN and SASE environments via recent enhancements to the Cloud EOS package, and its 7200 series switches are aimed at edge networking deployments.

Arista also this year rolled out WAN routing capabilities with new software, hardware and services, an enterprise-class system designed to link critical resources with core data-center and campus networks. That package, called the Arista WAN Routing System, ties together three components: enterprise-class routing hardware, software for its CloudVision management platform called Pathfinder, and the ability to establish neutral peering points called Transit Hubs. This enables customers to set up carrier-neutral and cloud-adjacent facilities to provide self-healing and path-optimization links across core, aggregation, and cloud networking interconnects.

Experts say the introduction of the WAN Routing package is significant because it represents Arista’s first official routing platform. In the past, Arista’s L2/3 data-center switches were deployed for routing use cases, but they were principally data center switches. Now, Arista is expressly targeting an expansive range of routing use cases with an unambiguous routing platform that specifically addresses the SD-WAN functionality gap in the Arista portfolio.

Making strides in security

Arista has never been considered a pureplay security vendor compared to Cisco, Juniper, Fortinet and others. However, in 2023 the vendor rolled out a SaaS-based service called CloudVision Guardian for Network Identity aimed at giving enterprises more network access control (NAC).

It uses real-time telemetry from Arista’s network products, combined with data from its CloudVision management platform, and takes advantage of AI to evaluate the information and implement security policies. The service can also onboard new devices, authenticate existing users, segment devices on the network, or help troubleshoot problems from a cloud-based system, Arista says.

In addition, late in 2023 the company partnered with zero-trust security vendor Zscaler to offer a cloud-based platform that promises to secure users, workloads, and devices over any network and location. The partnership also lets customers garner critical domain and attacker infrastructure intelligence for Arista NDR, which offers security diagnostics for the enterprise and lets organization react to threats more quickly.

Arista also offers a Macro Segmentation Service (MSS) that allows for the creation and enforcement of micro-perimeters through edge switches that can protect or isolate assets without requiring the deployment of firewalls across the enterprise network. Segmentation policies can be defined once in Arista CloudVision and enforced dynamically based on real-time network, application, device, or user identity information, according to Arista.

The company’s security strategy is nuanced, experts say, and nowhere is that more apparent than its zero-trust effort. “There’s a lot of vendors out there that say, ‘Buy our stuff, and that’s all you need.’ And I think, customers see through that,” said Rudolph Araujo, director of security for Arista.

“They realize that something like zero trust is much bigger than one vendor, or even any combination of vendors because it’s not just a technology problem. And so they like the fact that we’re a very open-standards-based ecosystem where anyone can integrate with us using OpenAPI-supported APIs,” Araujo said.

He adds that when it comes to security, Arista comes at it from the network perspective. “As we think about evolution, as we think about innovations, we’re always thinking about what that next frontier is, but in the context of the network, not, ‘Hey, we want to be everything to everyone.’ We know what we’re good at, and we’re going to stick with it,” Araujo said.

The opportunities are there

One of Arista’s greatest opportunities is growing its campus networking business, where the company is still looked at as a relatively new entrant, Boujelbene said.

Arista’s Hull agrees. “The opportunity in the campus is a significant future high-growth area for us, and while we have a good portfolio now, that’s not to say that there aren’t opportunities for that portfolio to continue to expand. Certainly 2024 and beyond, campus networks, especially around high-end and enterprise customers – that’s where our significant opportunity is going forward.”

Hull added that the midmarket or midsized enterprise campus is also a target area for growth going forward.

Arista has bolstered its data center and campus switching technologies in the past few years by snapping up software-defined capabilities from Pluribus and Big Switch in 2022 and 2020, respectively. The technology from those purchases continues to evolve in Arista’s portfolio, experts said.

“In the last three years, we have made an investment in and seen a significant uptick in enterprise customers wanting to do business with Arista,” Ullal told Wall Street analysts in a recent financial results call. “Historically, it’s been the high-tech enterprise and the financials and today we’re seeing a much better cross section of verticals, including healthcare and education.”

She added, “It’s pretty exciting that Arista is setting the tone for a two-tier defined architecture across the enterprise, and building that modern operating model based on CloudVision.”

Banking on generative AI

Beyond enterprise and campus evolution, Arista’s development of all things AI-networking, including the ability to support high-performance backend and front-end AI networks, will be another core driver for the company’s future.

“Everybody in the industry is talking about AI networking, and of course we have a good market position in high-end data center networking, but as the AI opportunity continues to expand and grow, clearly that’s an area for us to continue to be successful,” Hull said.

He adds, “Hand-in-hand with that development is the evolution and transition to 800G technology, and I have to anticipate at some point we will be talking about an 800G capable product.”

In early 2023, Arista said it was in the initial stages of developing an AI networking technology called AI Spine. Controlled by Arista EOS, it uses data-center switches with deep packet buffers and networking software to efficiently control AI traffic.

Arista’s AI Spine is based on its 7800R3 Series data-center switches, which, at the high end, support 460Tbps of switching capacity and hundreds of 40Gbps, 50Gbps, 100Gbps, or 400Gbps interfaces, along with 384GB of deep buffering. AI Spine systems would create high-bandwidth, lossless, low-latency, Ethernet-based networks that can ultimately interconnect thousands of GPUs at speeds of 100Gbps, 400Gbps, and 800Gbps, according to a white paper on AI Spine.

Arista is also a key member of the Ultra Ethernet Consortium (UEC), which is working to ensure that Ethernet technology will develop into the core connectivity layer for high-capacity AI processes. AMD, Arista, Broadcom, Cisco, Eviden, HPE, Intel, Meta and Microsoft make up the UEC.

AI networking will be about performance, and Arista has the kind of leaf-and-spine switches that will address that going forward, but AI networking is also about improving overall network operations to detect and prevent network problems before they occur. And that will drive interest from enterprises as well as hyperscalers, said Vijay Bhagavath, IDC vice president, cloud & datacenter networking research.

“Both Arista and Cisco have an opportunity to penetrate the backend network at the hyperscale and enterprise AI clouds,” Bhagavath said. “Over the next several years, we’ll see a lot more broad-based activity from enterprises, both training and fine-tuning generative models and then incorporating them into the line of business applications to streamline IT ops,” Bhagavath said.

Arista has already reported an uptick in AI networking development. It expects to see AI networking trials start in 2024, and production deployments become more widespread in 2025.