Cloud TPU

TPU VM exposing Google's ML accelerator chips, alone or in a multi-host slice.

Provider: gcp
Kind: Virtual machine
Category: Compute
Official docs: https://cloud.google.com/tpu/docs/intro-to-tpu

Properties · 4 set by this component

property	type	values	required	kind default	this component
chipsPerHost · component property	integer	—	no	8	inherited
diskEncryption	boolean	—	no	—	▸ true
instances	integer	—	no	—	inherited
managed	boolean	—	no	false	▸ false
memoryGib	integer	—	no	—	inherited
os	enum	linux \| windows \| aix \| ibm-i	no	linux	▸ linux
publicIp	boolean	—	no	—	inherited
sizeVcpu	integer	—	no	—	inherited
sliceScope · component property	enum	single-host \| multi-host	no	single-host	inherited
tpuVersion · component property	enum	v4 \| v5e \| v5p \| v6e \| tpu7x	no	—	inherited
zoneRedundant	boolean	—	no	—	▸ false

Rules that apply · 1

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.

Stateful node without zone redundancyerror

stateful-node-without-zone-redundancy

A stateful compute node pinned to exactly one availability zone disappears with that zone, and its state goes with it. Nodes inside a cell are exempt, because there the cell is deliberately single-zone and redundancy comes from the other cells.

— all of these hold:
— the node's `stateless` is false
— the node sits inside exactly 1 availability-zone group
— this does not hold:
— the node sits inside at least one cell group

Use it in a design

nodes:
  - { id: "cloud-tpu", component: "google-cloud-tpu" }

Other Virtual machine products · 60

Auto Scalingalibaba-auto-scalingalibabacloudvirtual-machine Elastic Compute Servicealibaba-ecsalibabacloudvirtual-machine Edge Node Servicealibaba-ensalibabacloudvirtual-machine Simple Application Serveralibaba-simple-application-serveralibabacloudvirtual-machine Amazon EC2aws-ec2awsvirtual-machine Amazon EC2 Auto Scalingaws-ec2-auto-scalingawsvirtual-machine Amazon EC2 Dedicated Hostsaws-ec2-dedicated-hostawsvirtual-machine Amazon Elastic VMware Serviceaws-evsawsvirtual-machine Amazon Lightsailaws-lightsailawsvirtual-machine AWS Nitro Enclavesaws-nitro-enclaveawsvirtual-machine Azure BareMetal Infrastructureazure-baremetal-infrastructureazurevirtual-machine Azure Compute Fleetazure-compute-fleetazurevirtual-machine Azure Dedicated Hostazure-dedicated-hostazurevirtual-machine Azure Data Science Virtual Machineazure-dsvmazurevirtual-machine Nutanix Cloud Clusters on Azureazure-nutanix-cloud-clustersazurevirtual-machine Azure Virtual Machinesazure-virtual-machineazurevirtual-machine Azure Virtual Machine Scale Setsazure-vm-scale-setazurevirtual-machine Azure VMware Solutionazure-vmware-solutionazurevirtual-machine Crusoe Cloud VMcrusoevirtual-machine Firecrackerfirecrackervirtual-machine Microsoft Hyper-Vhyper-vvirtual-machine KubeVirtkubevirtvirtual-machine Lambda Cloud instancelambda-labsvirtual-machine KVM with libvirtlibvirtvirtual-machine Nebius AI Cloud VMnebiusvirtual-machine Nutanix AHVnutanix-ahvvirtual-machine OpenStack Novaopenstack-novavirtual-machine oVirtovirtvirtual-machine Proxmox VEproxmoxvirtual-machine Runpod Podrunpodvirtual-machine Vast.ai instancevast-aivirtual-machine Generic virtual machinevirtual-machinevirtual-machine VMware vSpherevspherevirtual-machine XCP-ngxcp-ngvirtual-machine XenServerxenservervirtual-machine DigitalOcean Bare Metal GPUsdigitalocean-bare-metal-gpudigitaloceanvirtual-machine Dropletsdigitalocean-dropletdigitaloceanvirtual-machine Droplet Autoscale Pooldigitalocean-droplet-autoscaledigitaloceanvirtual-machine Bare Metal Solutiongoogle-bare-metal-solutiongcpvirtual-machine Compute Enginegoogle-compute-enginegcpvirtual-machine

and 20 more → Virtual machine page

Source

1# yaml-language-server: $schema=../../schema/component.schema.json
2# Cloud TPU: Google's ML accelerator ASICs, reached as TPU VMs built on Compute Engine
3# ("Introduction to Cloud TPU"). Chips are grouped into slices; on v6e a slice of up to 8 chips
4# attaches to a single host and 4x4 and larger topologies need multiple hosts, up to a 256-chip pod
5# ("TPU v6e"). Code must be compiled by XLA.
6schemaVersion: "1"
7type: "component"
8id: "google-cloud-tpu"
9kind: "virtual-machine"
10provider: "gcp"
11docs: "https://cloud.google.com/tpu/docs/intro-to-tpu"
12defaults:
13  managed: false
14  os: "linux"
15  diskEncryption: true
16  zoneRedundant: false
17properties:
18  tpuVersion:
19    type: "enum"
20    values: ["v4", "v5e", "v5p", "v6e", "tpu7x"]
21  sliceScope:
22    type: "enum"
23    values: ["single-host", "multi-host"]
24    default: "single-host"
25  chipsPerHost:
26    type: "integer"
27    default: 8

The same page as markdown, for an agent: /catalogue/components/google-cloud-tpu.md