Skip to content
In development

Somewhere to run the heavy work

Planned to schedule GPU work for Ethen products from a single description of each job.

GPU Compute is the Ethen Platform subsystem planned to schedule GPU workloads. Today a facade and workload templates exist; integration with a serving engine is in development.

Check current statusFaros research

Compute is part of Ethen Platform. Its facade is the entry point the Platform plans to use for GPU work, so products would not call a GPU provider directly.

Illustration

No GPU needed · 4 steps

Runs on the CPU worker that exists in development.

  1. 01Product. An Ethen product describes the work it needs. Products would not call a GPU provider directly.
  2. 02Compute facade. A compute facade: the interface the Platform uses to describe a workload.
  3. 03Template match. Workload templates that describe common job shapes.
  4. 04CPU worker. A CPU worker for jobs that do not need a GPU.
  • Exists in development
  • Planned
  • Not integrated
Illustration of the intended design, not a capture. No GPU service is running for customers.

01Facade

A workload, described once

A product would describe its workload to the facade, which matches it to a template. The two shapes below are written for this page: they are not the facade's schema, and there is no public CLI or SDK.

01A product describes a workload

Illustration
yaml
# Illustrative, for this page
workload:
  product: <ethen-product>
  template: <template-name>
  needs_gpu: false
Illustration: a shape written for this page, not a real schema

02The facade answers with a route

Illustration
yaml
# Illustrative, for this page
route:
  template: <template-name>
  runs_on: cpu-worker     # exists in development
  gpu_engine: not-integrated
Illustration: a shape written for this page, not a real schema

02Specification

Not published

We publish no prices, locations, hardware types, endpoints or service-level commitments.
Check current status
ItemPublished value
CapacityNot publishedNo GPU service is running for customers.
Hardware typesNot publishedNo hardware type is offered.
LocationsNot publishedNo location is offered.
PricesNot publishedYou cannot rent GPUs from us.
Service-level commitmentsNot publishedNone is offered.
EndpointsNot publishedNo endpoint is public.
StatusIn developmentFrom the status registry.

03What exists

A facade, templates and a CPU worker

Three parts exist in development. Integration with a GPU serving engine does not, and engines that are selected but not integrated are not named here.
  • 01Exists in development

    Compute facade

    A compute facade: the interface the Platform uses to describe a workload.

  • 02Exists in development

    Workload templates

    Workload templates that describe common job shapes.

  • 03Exists in development

    CPU worker

    A CPU worker for jobs that do not need a GPU.

The three covers are abstract illustrations.

Planned

  • Integration with an open-source GPU serving and scheduling engine. Engines have been selected but are not integrated, so we do not name them here.
  • Scheduling GPU workloads for Ethen products through the facade.

04Limits

No GPU service, and no rental

Compute is in development. The Faros research pilot is a separate activity, described on its own page.
Faros research
  • No GPU service is running for customers, and you cannot rent GPUs from us.
  • We publish no prices, locations, hardware types, endpoints or service-level commitments.
  • The Faros research pilot used a rented GPU. That was a separate research activity, not Platform Compute.
  • Not available for general use; nothing here is a release promise or date.

05Questions

Questions

  1. 01Can I rent GPUs today?

    No. GPU Compute is in development. Only a facade, templates and a CPU worker exist, and no GPU service runs for customers.

  2. 02Which GPU engine does it use?

    None is integrated yet. Open-source engines have been selected; we will name them on this page once they are integrated.

  3. 03Why does this page say “Not published” so often?

    Because nothing has been published. We publish no prices, locations, hardware types, endpoints or service-level commitments, and no GPU service is running for customers.

Status and related pages

GPU Compute is part of Ethen Platform and is in development. No GPU service is running for customers.

See availability