Arista Foundations key terms

Arista's network operating system, built as a single modular software image that runs across its switch families. It provides a Linux-based environment with a consistent CLI, programmatic interfaces, and in-service software upgrades.

A central in-memory database inside EOS that holds the state of all system components. Processes read and write state through it, so a failure in one agent does not crash the whole system.

An EOS capability where a failed software process restarts and repopulates its state without reloading the switch or dropping forwarding, keeping traffic flowing through the data plane.

A method of upgrading EOS software while the switch keeps forwarding traffic, using process-level restarts and redundancy rather than a full reload that interrupts service.

A management and automation platform that provides a single view of the network, configuration and image management, change workflows, and telemetry collection across Arista devices.

The web-based user interface and workflow layer of CloudVision where operators stage changes, review compliance, view topology, and push configurations to devices.

The on-premises controller component of CloudVision that handles device connectivity, state streaming, and coordination services between network devices and the management platform.

A cloud-hosted delivery model of CloudVision in which Arista operates the management plane, and devices connect to it over the internet through a secure tunnel.

An EOS agent that streams telemetry, configuration, and state data from the switch to CloudVision, enabling visibility and management without polling devices directly.

An open gRPC-based protocol for reading, writing, and subscribing to device configuration and state. It is commonly used for streaming telemetry and model-driven management.

A vendor-neutral set of data models for network device configuration and state, allowing tools to manage devices from different vendors with consistent schemas.

A CloudVision workflow that bundles configuration changes into a reviewable, approvable unit. It lets teams validate, schedule, and roll back changes with an audit trail.

The practice of comparing running device configurations against a defined intended state or policy, then flagging and remediating drift from that baseline.

A two-tier data center design where every leaf switch connects to every spine switch. Traffic between leaves crosses one spine hop, giving predictable latency and easy horizontal scaling.

The upper tier of a leaf-spine fabric. Spines interconnect leaves and typically do not connect to servers or endpoints, keeping the fabric backbone simple and high-bandwidth.

The lower tier of a leaf-spine fabric that connects servers, storage, and edge devices to the network and uplinks to every spine switch.

A routing technique that load-balances traffic across multiple equal-cost paths to the same destination, using a hash of packet header fields to keep flows consistent.

Multi-Chassis Link Aggregation, an Arista feature that lets two switches act as one logical device for link aggregation, providing redundant server or leaf connections without spanning tree.

A bundle of physical links treated as one logical link to increase bandwidth and provide redundancy. Traffic is distributed across member links by a hash.

Link Aggregation Control Protocol, an IEEE standard that negotiates and monitors which physical links belong to a link aggregation group, detecting misconfigurations and failures.

A network virtualization encapsulation that carries Layer 2 frames inside UDP packets across a Layer 3 network, allowing large numbers of isolated virtual networks to span a data center fabric.

The device or interface that encapsulates and decapsulates VXLAN traffic, mapping local Layer 2 segments into VXLAN network identifiers and forwarding them across the underlay.

Ethernet VPN, a BGP-based control plane that distributes MAC and IP reachability information for overlay networks, enabling efficient forwarding and multi-homing in VXLAN fabrics.

A path-vector routing protocol used to exchange reachability information between autonomous systems and increasingly as the routing protocol inside large data center fabrics.

A number identifying a routing domain that exchanges routes using BGP. In data centers, each leaf and spine often gets its own private ASN for scalable fabric routing.

External BGP, the form of BGP used between routers in different autonomous systems. In leaf-spine fabrics it is often used between every leaf and spine to build an ECMP underlay.

Open Shortest Path First, a link-state interior gateway protocol that builds a topology map and computes shortest paths using cost, commonly used in enterprise and underlay routing.

Combining multiple specific prefixes into a single less-specific route advertisement, which reduces routing table size and hides topology changes from other parts of the network.

Continuous, periodic push of device state and counters to a collector rather than polling by SNMP. It gives finer-grained, near-real-time visibility with lower device overhead.

Simple Network Management Protocol, a traditional polling-based protocol for monitoring and managing network devices through structured management information base objects.

A packet-sampling technology that sends sampled packet headers and interface counters to a collector, giving visibility into traffic flows with low overhead.

Traffic analysis that aggregates packet metadata into flow records, showing who is talking to whom, on which ports and protocols, for capacity and security analysis.

An ordered set of rules that permits or denies traffic based on fields such as source, destination, protocol, and port. It is applied to interfaces or control-plane policies.

A mechanism that rate-limits traffic destined to the switch's own control plane, protecting routing protocols and management access from being overwhelmed by floods or attacks.

An IEEE 802.1AE standard for encrypting and authenticating Ethernet frames on a link, providing hop-by-hop confidentiality and integrity for data in transit.

A security model that assumes no implicit trust based on network location, requiring identity verification and least-privilege access for every user, device, and workload.

Managing devices through structured data models and APIs instead of screen-scraping the CLI, enabling automation, validation, and consistent configuration across many devices.

Arista's JSON-based REST API for EOS, letting automation tools issue commands and retrieve structured output over HTTP or HTTPS instead of parsing CLI text.