name: Quickwit FinOps
description: Cost and financial operations considerations for running Quickwit, a cloud-native open-source search engine designed to minimize infrastructure spend by decoupling compute from storage.
url: https://quickwit.io/
cost_model:
  type: self-hosted open source
  license: Apache 2.0
  software_cost: 0
  primary_cost_drivers:
    - name: Object Storage
      description: >
        Quickwit stores indexes on cloud object storage (Amazon S3, Google Cloud
        Storage, Azure Blob Storage). Storage costs are determined by the cloud
        provider's object-storage pricing and the total compressed index size.
        Because Quickwit uses columnar compression, index storage is typically a
        small fraction of raw log volume.
      typical_unit: GB/month
      examples:
        - provider: Amazon S3
          cost_per_gb_month: ~0.023 USD (standard tier, us-east-1)
        - provider: Google Cloud Storage
          cost_per_gb_month: ~0.020 USD (standard tier)
    - name: Compute (Indexers and Searchers)
      description: >
        Quickwit separates indexing and searching into stateless worker roles.
        Indexers can be scaled down or stopped when ingestion is idle, and
        searchers handle query traffic on demand. Users pay only for the
        compute instances they provision (EC2, GKE nodes, bare metal, etc.).
      typical_unit: vCPU-hour
    - name: Data Transfer / Egress
      description: >
        Queries pull index data from object storage into searcher memory.
        Cloud providers charge for data transfer between storage and compute
        in certain configurations. Keeping searchers co-located with the
        storage bucket's region minimizes egress fees.
      typical_unit: GB transferred
cost_optimization:
  - tip: Use object-storage tiering (e.g., S3 Intelligent-Tiering) for cold index data.
  - tip: Scale indexer nodes to zero during off-peak hours; Quickwit's stateless design supports this.
  - tip: Tune retention policies to automatically delete old splits and reclaim storage costs.
  - tip: Leverage Quickwit's schemaless compression to reduce stored index size versus Elasticsearch.
  - tip: Run Quickwit on spot or preemptible instances for indexer workloads to cut compute costs significantly.
  - tip: Use the "auto" commit mode for ingest to batch writes efficiently, reducing API overhead.
references:
  - title: Quickwit Architecture (decoupled compute/storage)
    url: https://quickwit.io/docs/overview/architecture
  - title: How Binance built a 100PB log service with Quickwit
    url: https://quickwit.io/blog/binance-one-hundred-pb-log-service
  - title: Quickwit Kubernetes Helm Deployment
    url: https://quickwit.io/docs/deployment/kubernetes/helm
notes: >
  Quickwit joined Datadog in January 2025. The open-source codebase remains
  active under Apache 2.0. For enterprises requiring a managed service with
  predictable cost, third-party managed hosting (e.g., Elestio) is available
  at a per-node fee. Contact Datadog for enterprise pricing inquiries.