specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Ensembl
providerId: ensembl
created: '2026-06-13'
modified: '2026-06-13'
reconciled: true
tags:
  - Genomics
  - Bioinformatics
  - Genome Annotation
  - Life Sciences
  - FinOps
  - FOCUS
description: >-
  FOCUS-aligned FinOps for Ensembl. The Ensembl REST API carries zero direct
  monetary cost — it is a free public scientific resource operated by
  EMBL-EBI and Wellcome Sanger Institute. FinOps considerations therefore
  focus on the indirect cost of compute consumed while calling the API
  (cloud egress, worker time, retry overhead) and on choosing the correct
  data access method (REST API vs. FTP bulk download vs. local database
  mirror) to minimise wasted spend in downstream pipelines. Pipelines that
  exceed the 55,000 req/hr rate limit incur opportunity cost from retries
  and pipeline delays rather than direct API charges.
sources:
  - https://rest.ensembl.org
  - https://github.com/Ensembl/ensembl-rest/wiki/Rate-Limits
  - https://www.ensembl.org/info/docs/api/index.html
alignedWith:
  framework: FinOps Foundation Framework
  frameworkUrl: https://www.finops.org/framework/
  dataSpec: FOCUS
  dataSpecVersion: '1.3'
  dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/
publisherName: EMBL-EBI / Wellcome Sanger Institute
serviceCategory: Bioinformatics / Genomics Data
billingModel:
  pricingCategory: Free (No Direct Charge)
  billingFrequency: N/A
  billingCurrency: N/A
  chargeCategories:
    - Usage
focusColumns:
  ServiceName: Ensembl REST API
  ServiceCategory: Bioinformatics
  ProviderName: Ensembl (EMBL-EBI)
  PublisherName: European Bioinformatics Institute
  InvoiceIssuerName: N/A
  BillingCurrency: N/A
  ChargeCategory: Usage
meters:
  - name: api_requests
    description: >-
      Count of HTTP requests made to rest.ensembl.org. No direct charge, but
      each request counts against the 55,000 req/hr quota and consumes
      downstream compute time (worker CPU, network egress).
    unit: request
    aggregation: sum
    dimensions:
      - endpoint_category
      - method
      - species
  - name: retry_requests
    description: >-
      Count of retry requests triggered by 429 rate limit responses. Each
      retry represents wasted pipeline time and indirect compute cost.
    unit: request
    aggregation: sum
    dimensions:
      - endpoint_category
      - pipeline_name
  - name: bulk_download_bytes
    description: >-
      Volume of data retrieved via Ensembl FTP bulk download as an alternative
      to repeated REST calls. Relevant for cloud egress cost attribution.
    unit: byte
    aggregation: sum
    dimensions:
      - dataset
      - species
      - release_version
principles:
  - name: Visibility
    description: >-
      Instrument pipeline code to count REST API calls per endpoint category
      and species. Track X-RateLimit-Remaining in logs to detect when
      pipelines are approaching quota exhaustion before 429s appear.
  - name: Allocation
    description: >-
      Tag API calls with a pipeline_name or team label in application logs so
      that indirect compute cost (worker time, egress) can be attributed to
      the correct cost centre. Separate test and production pipeline traffic
      by IP or via local mirrors to prevent development runs from consuming
      the shared production quota.
  - name: Optimization
    description: >-
      Use batch POST endpoints for multi-identifier lookups to reduce per-call
      overhead. For full-genome or species-wide datasets, prefer the Ensembl
      FTP service over repeated REST calls to avoid rate limit delays. Stand
      up a local Ensembl database mirror for high-throughput institutional
      pipelines that regularly exceed 55,000 req/hr. Cache stable reference
      data (gene annotations, assembly metadata) locally to eliminate
      redundant API calls.
  - name: Accountability
    description: >-
      Assign a bioinformatics lead as the responsible owner for API usage
      budgets within shared compute environments. Review pipeline retry rates
      monthly; a rising retry rate signals that the pipeline needs batching
      optimisation or a local mirror before cloud egress costs escalate.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com