OpenStack Cyborg
Accelerator management framework that discovers FPGAs, GPUs, NICs, SSDs and AI chips, reports them to Placement and attaches them to instances.
- Provider
- openstack
- Category
- Operations
- Official docs
- https://docs.openstack.org/cyborg/latest/
- Also known as
- Accelerator Life Cycle Management
| property | type | values | required | kind default | this component |
|---|---|---|---|---|---|
| acceleratorType | enum | fpga | gpu | nic | ssd | ai-chip | no | — | inherited |
| consensus | enum | none | raft | paxos | zab | no | none | inherited |
| deviceProfile | boolean | — | no | true | inherited |
| managed | boolean | — | no | false | ▸ false |
| nodes | integer | — | no | — | inherited |
| placementInventory | boolean | — | no | true | inherited |
| replicated | boolean | — | no | false | inherited |
| role | enum | metadata | lock | discovery | placement | sharding | leader-election | config | yes | — | ▸ placement |
Every rule whose selector covers this component's kind or its category. Whether a rule fires depends on the design: its properties, its edges and its groups.
shared-dependency-without-redundancy
The design requires redundant discovery, but this DNS or coordination service, which everything depends on, runs on a single provider or a single replica. Its failure is the whole system's failure.
- — all of these hold:
- — the design requirement `availability.discoveryRedundant` is true
- — any of these hold:
- — the node's `redundantProviders` is false
- — the node's `replicated` is false
stateful-engine-pinned-to-single-zone
A matching engine, coordination service or event store living in exactly one availability zone, with no failover and no replication, is lost with that zone. Everything that depends on it stops with it.
- — all of these hold:
- — the node sits inside exactly 1 availability-zone group
- — any of these hold:
- — the node's `failover` is "none"
- — the node's `replicated` is false
- — the node's `replication` is "none"
nodes:
- { id: "cyborg", component: "openstack-cyborg" }Microservices Enginealibaba-mseAWS Cloud Mapaws-cloud-mapAzure Arcazure-arcAzure Device Registryazure-device-registryDurable Objectscloudflare-durable-objectsApache Helixapache-helixConsulconsulGeneric coordination servicecoordination-serviceetcdetcdEurekaeurekaNacosnacosOxiaoxiaApache ZooKeeperzookeeperService Directorygoogle-service-directoryEndpointsk8s-endpointsOpenStack Blazaropenstack-blazarOpenStack Placementopenstack-placement
# yaml-language-server: $schema=../../schema/component.schema.json # OpenStack Cyborg: general management framework for accelerators. Per "OpenStack Accelerator (Cyborg)" the # cyborg-agent discovers local FPGAs, GPUs, NICs, SSDs and AI chips via drivers, reports inventory to the # conductor for Placement, and manages device programming and attachment to Nova instances through device # profiles: a placement role over the cloud, with no consensus protocol of its own. schemaVersion: "1" type: "component" id: "openstack-cyborg" kind: "coordination-service" provider: "openstack" docs: "https://docs.openstack.org/cyborg/latest/" aliases: ["Accelerator Life Cycle Management"] defaults: role: "placement" managed: false properties: acceleratorType: type: "enum" values: ["fpga", "gpu", "nic", "ssd", "ai-chip"] placementInventory: type: "boolean" default: true deviceProfile: type: "boolean" default: true