Cloudvision APIs are state based, resource-oriented APIs modeled in Protobuf and accessed over gRPC using a standardized set of RPC verbs.
CloudVision is a powerful platform that processes and stores tremendous amounts of network data. It knows the topology of the network, device configuration, interface activity and other network events. These APIs allow access to fleet-wide data access and control, forming a management-plane with consistent usage.
Functionality is defined in a data-oriented (rather than action-oriented) form. Designing the APIs to use state-synchronization confers some desirable traits:
For more information on modelling, see the Modeling page.
Adding a new API simply means creating a new model. Once a model is made, a set of consistent API verbs can then be generated for it. These verbs are named and behave the same across all models. For clients this makes it near trivial to adapt usage of one API to another.
Synchronous APIs, while simple, have some classic problems. Take for example a DoReboot
operation.
This operation has numerous failure modes:
In each of these cases, the client is now responsible of determining which failure mode occurred and how to proceed.
In a state-sharing paradigm a reboot request might be performed by setting a reboot request attribute to
the current time. The service can return from this request almost immediately (only durably writing the request).
Then the client can subscribe to a last-reboot
attribute and knows the device has been rebooted successfully
when that attribute’s timestamp exceeds that of the reboot request. State sharing allows the various components
involved in an action not to need to care about each other; they just need to synchronize state when they
come up, and then they’ll do the right thing.
By modelling in protobuf
and exposing gRPC
RPCs, this data and management is accessible from nearly any
environment in nearly any language.
For Arista-supported clients, see the Existing Clients page.
Alternatively, a basic guide on creating your own.
In addition to pure-gRPC clients, HTTP REST mappings are easily generated through the great gRPC ecosystem. Thus, APIs can be utilized from nearly anywhere.
Another benefit of using protobuf
and gRPC
is the vast ecosystem surrounding them.
While portable client-generation is useful a well-supported ecosystem provides even more tooling.
This list is not extensive, nor an endorsement of any project. Many more tools can be found on the awesome-grpc page.