Highlights:

  • CloudVision is powered by Arista’s Network Data Lake, a centralized data repository that gathers and unifies data from across the network.
  • In data centers, CloudVision provides real-time insights and automation, using CV UNO’s machine learning to detect and resolve issues.

Since its founding in 2004, Arista Networks Inc. has built a reputation for producing top-tier network switches driven by high-quality hardware and a unified, open operating system across all its products.

In 2015, Arista Networks introduced “CloudVision,” shifting its management, orchestration, and automation features to the cloud. Traditionally, CloudVision played a secondary role in Arista’s network infrastructure, which has been the company’s primary differentiator since its inception. This week, Arista introduced several updates to the CloudVision platform, improving its usability for businesses to manage networks across data centers, campuses, and cloud environments.

Organizations often struggle to manage various parts of their network using distinct tools, technologies, and operational models. For instance, an IT team may rely on one set of tools for managing data center operations, another for overseeing campus networks, and yet another for cloud services.

This fragmented approach has led to increased operational costs and complexity, making it more difficult to coordinate tasks, troubleshoot problems, and maintain consistent security across the organization. Additionally, it has complicated the processes of integrating, expanding, and upgrading the network.

CloudVision seeks to eliminate silos by consolidating network management into a single platform. Initially built to meet the stringent requirements of cloud service providers, CloudVision now offers these advantages to businesses of all sizes. What sets CloudVision apart is its ability to maintain “feature parity” with Arista’s Extensible Operating System (EOS), meaning it supports all the same features as EOS without the need for frequent updates.

In conventional network management systems, updates to network operating systems often lead to an 18 to 24-month delay before management tools can catch up. With CloudVision, however, any new feature added to EOS is immediately supported, allowing enterprises to leverage the latest network capabilities without delay.

CloudVision runs on Arista’s Network Data Lake, a centralized system that collects and consolidates data from across the network. With AI-driven analytics and predictive insights, network operators can detect and address potential issues before they become critical.

“Our enterprise approach is to build the best-in-class network. Our open standards, design principles and high availability capabilities — all allow us to do that. On the operations side, this is where CloudVision comes in, we focus on simplifying and reducing the headcount needed to operate the network. Then, of course, security has become a big part of our story as well,” said Vice President of EOS product management and services at Arista, Jeff Raymond.

Here’s a detailed overview of how CloudVision supports various aspects of network management:

In data centers, CloudVision offers real-time insights and automation to simplify network management. Its Universal Network Observability (CV UNO) leverages machine learning to identify and resolve issues. CV UNO provides a comprehensive view of network and application data, illustrating how different components of the infrastructure interact, enabling quicker troubleshooting and more efficient management.

In campus environments, which typically involve a wider range of users, CloudVision streamlines operations by providing guided workflows for network provisioning, management, and troubleshooting. The platform encompasses all aspects from day-zero setup to day-one and day-two operations, including tasks like making port changes, adding or removing networks, and conducting diagnostics.

A standout feature in campus networks is the Ask AVA autonomous virtual assistant. This generative AI chatbot is integrated into CloudVision, enabling operators to communicate with the platform using natural language. This feature makes it easier for non-expert users, like help desk staff and junior network administrators, to navigate the system.

For wide-area networks, CloudVision offers operator dashboards that display WAN connections through telemetry, allowing for network health monitoring. A key feature of WAN management is geographic mapping, which shows the locations and traffic patterns of WAN links. The platform utilizes dynamic routing, leveraging real-time data to select the most efficient paths for application traffic. Additionally, it integrates with Google Cloud Platform and Amazon Web Services, delivering a unified view of both on-premises and cloud-based WAN networks.

“The key to our approach is we have this much broader data set, especially with our cloud service offering, where it’s global and anonymized. So we can start to train data and look at a much bigger sample size. This is just the start of what we’ve built into CloudVision from a generative AI perspective,” said Raymond.

Finally, CloudVision improves zero-trust security through multidomain segmentation. It offers orchestration and monitoring services for zero-trust policies, enabling organizations to enforce security rules based on micro-perimeters. These policies are overseen through dedicated security dashboards, enabling network operators to visualize and manage traffic flow across the enterprise easily. CloudVision integrates with firewalls like Palo Alto Networks, ServiceNow for IT management, and VMware for virtualization, offering companies a complete perspective on their network security.

A comprehensive view enables organizations to streamline the management of hybrid environments and ensure consistency across the network. Many enterprises have already implemented CloudVision, initially in data centers and later expanding to campus networks, WAN, and security environments. CloudVision has evolved into a core component of Arista. When customers choose Arista, CloudVision is automatically included as part of the complete solution rather than being an optional add-on.

“We have been late to all the markets we’ve entered, and we’ve been able to build the product correctly. It ends up playing out well in the long run. That applies to EOS, our network data lake that we continue to invest in, and CloudVision at the application level,” Raymond said.

For Arista, the emphasis on CloudVision is encouraging. Its growth has been largely fueled by its top-tier hardware and software, with the cloud enhancing the overall strategy. We are swiftly entering the era of AI in networking, presenting a disruptive opportunity for the company. According to my recent survey on AI networking, an impressive 93% of customers would consider switching vendors if the AI capabilities are significantly superior.

AI excellence will be provided through the cloud, and although CloudVision is a robust platform, there is currently no definitive leader in AI for networking. Raymond emphasized that there are more developments ahead for CloudVision, as Arista aims to utilize the cloud for future innovations.