Arista CloudVision: Network-Wide Automation and Telemetry
Modern networks are growing faster than human administrators can manage them manually. Configuration drift, slow troubleshooting cycles, and siloed visibility are becoming critical bottlenecks for enterprises running complex multi-site infrastructure. Arista CloudVision was designed to solve exactly these problems, providing a centralized platform that brings together network-wide automation, real-time telemetry, and change management across every device running Arista EOS.
Whether you are managing a handful of switches in a data center or thousands of devices spread across multiple campuses, CloudVision offers a unified operational model that transforms how teams deploy, monitor, and maintain their networks.
What Is Arista CloudVision?
Arista CloudVision (CVP) is a network-wide automation and telemetry platform built specifically for infrastructure running Arista EOS. It acts as a centralized management plane that integrates deeply with every EOS device through a persistent streaming telemetry connection.
Unlike traditional network management systems that rely on periodic polling with SNMP, CloudVision establishes a continuous, bidirectional channel to each device. This means state changes, counters, routing table updates, and hardware telemetry are streamed in real time to the platform, giving operators an always-current view of their entire network.
CloudVision is available in two deployment models:
- CVP On-Premises: A cluster of physical or virtual appliances deployed inside your own data center
- CVX Cloud (CVaaS): A cloud-hosted version of CloudVision delivered as a SaaS offering, removing the burden of managing the platform infrastructure itself
Both options share the same core capabilities, so teams can choose based on their operational and compliance requirements.
The Role of Arista EOS in the CloudVision Ecosystem
You cannot fully appreciate CloudVision without understanding the role of Arista EOS. EOS is the single, consistent operating system that runs on every Arista switch and router. One of its defining architectural decisions is that the entire system state — running processes, hardware tables, interface counters, BGP neighbors, MLAG state — is maintained in a central in-memory database called SysDB.
This architecture makes EOS inherently programmable and telemetry-friendly. CloudVision connects to each device using eAPI and the OpenConfig/gNMI streaming telemetry stack, pulling structured data directly from SysDB without needing to screen-scrape CLI output or rely on unreliable SNMP traps.
The result is a tight integration between EOS and CloudVision that produces:
- Sub-second telemetry latency for key metrics
- Structured, schema-consistent data across all device types
- Reliable configuration push and validation at scale
This is not a bolt-on management tool — it is a purpose-built ecosystem where the device OS and the management platform were designed to work together.
Network-Wide Automation with CloudVision
Configlet-Based Configuration Management
CloudVision manages device configurations through a concept called configlets — reusable, modular configuration snippets that can be applied to individual devices, containers of devices, or the entire network hierarchy.
For example, you might create a configlet for your NTP and DNS settings that applies globally, another for your standard BGP policy that applies to all spine switches, and device-specific configlets for unique interface configurations. CloudVision merges these into a complete intended configuration for each device.
This layered approach gives teams several practical advantages:
- Consistency: A change to a shared configlet automatically propagates to every device that references it
- Traceability: You know exactly which configlet contributes which lines to a device’s running config
- Scalability: Adding a new device means assigning it to the right containers — the correct configuration follows automatically
Change Control and Rollback
One of the most operationally valuable features in CloudVision is its Change Control workflow. Before any configuration change is pushed to production devices, it goes through a structured review and approval process built directly into the platform.
A change control record captures:
- Which devices are being changed
- What the before and after configuration looks like (a full diff)
- Who approved the change and when
- Whether any associated snapshot or task validation passed
This gives teams a built-in audit trail without relying on external ticketing integrations. If something goes wrong after a deployment, CloudVision stores complete snapshots of each device’s state before and after the change, enabling precise rollback to a known-good configuration with a single action.
Automated Provisioning with ZTP
CloudVision integrates with Zero Touch Provisioning (ZTP) to fully automate the bring-up of new EOS devices. When a new switch powers on and reaches the network, it contacts CloudVision, which identifies the device by its serial number or MAC address, assigns it to the correct container in the hierarchy, and automatically pushes the appropriate configlets.
A device that previously required a technician to manually cable, log in, and type configuration commands can now be fully operational within minutes of being racked — with no human interaction beyond physical installation.
Real-Time Telemetry and Streaming Analytics
Network Data Lake
CloudVision functions as a network data lake, continuously ingesting streaming telemetry from every EOS device and storing it in a time-series database. This historical record enables a capability Arista calls Network Time Machine.
With Network Time Machine, operators can rewind the state of any device or the entire network to any point in the past. If a user reports that their application was slow three hours ago, you can literally travel back in time within CloudVision and see exactly what the routing tables, interface utilization, and BGP state looked like at that specific moment.
This capability fundamentally changes troubleshooting. Instead of piecing together syslog fragments and hoping someone ran a show command at the right time, you have a complete, structured, queryable record of network state history.
Streaming Telemetry Metrics
CloudVision collects a broad range of telemetry data, including:
- Interface statistics: bandwidth utilization, error counters, packet drops
- Control plane state: BGP session status, OSPF adjacencies, VXLAN tunnel health
- Hardware tables: TCAM utilization, MAC table size, ARP/ND table entries
- System health: CPU usage, memory pressure, temperature sensors, power supply status
- Security events: ACL hit counts, traffic anomalies
All of this data is visualized in built-in dashboards and is also available via a REST API, allowing teams to integrate CloudVision telemetry into external platforms like Grafana, Splunk, or custom observability stacks.
Topology and Path Visualization
CloudVision automatically builds and maintains a live network topology map based on LLDP data streamed from devices. This map is not static documentation — it updates in real time as links come up and go down, giving operators an accurate picture of how the network is physically connected at any moment.
On top of this topology, CloudVision offers multi-layer path visualization, allowing engineers to trace how traffic flows from source to destination across the network. This is particularly powerful in VXLAN/EVPN fabrics, where the overlay and underlay can make traffic paths difficult to follow manually.
Integration With the Broader Ecosystem
CloudVision does not operate in isolation. Arista has invested heavily in making it a composable platform that fits into existing operational toolchains.
APIs and Programmability
Every CloudVision function is exposed through a well-documented REST API, making it straightforward to integrate with:
- CI/CD pipelines: Trigger change control workflows from a Jenkins or GitHub Actions pipeline after code review approval
- ITSM tools: Sync change records with ServiceNow or Jira automatically
- Ansible and Terraform: Arista publishes official modules for both, allowing infrastructure-as-code practitioners to manage CloudVision objects declaratively
CloudEOS and Multi-Cloud
For organizations with workloads running in public cloud environments, Arista offers CloudEOS — a software version of EOS that runs as a VM or container in AWS, Azure, and Google Cloud. CloudEOS devices connect to CloudVision just like physical hardware, extending the same automation and telemetry model into multi-cloud and hybrid architectures.
This consistency is a significant operational benefit. Your teams use the same tools, the same workflows, and see the same telemetry format regardless of whether the “switch” is a physical 7050 in your data center or a CloudEOS instance running in a VPC.
Common Use Cases
To make this concrete, here are some real-world scenarios where CloudVision delivers clear value:
- Leaf-Spine Fabric Management: Automate the deployment and ongoing management of VXLAN/EVPN data center fabrics at scale
- Campus Modernization: Extend zero-touch provisioning to branch and campus switches, eliminating pre-staging work
- Compliance Auditing: Use configuration snapshots and change history to demonstrate network state to auditors without manual evidence collection
- Proactive Capacity Planning: Analyze historical interface utilization trends to identify congestion risks before they become incidents
- Security Operations: Feed CloudVision telemetry into SIEM platforms for network-based threat detection
Conclusion
Arista CloudVision represents a modern approach to network management, one where automation is not an add-on feature but a foundational design principle. By combining the programmable architecture of EOS with network-wide streaming telemetry, structured change management, and deep API integration, CloudVision gives network teams the visibility and control they need to operate at the speed modern applications demand.
For organizations already running Arista EOS, deploying CloudVision is one of the highest-leverage investments available — turning a collection of well-configured devices into a fully observable, consistently automated network infrastructure.
