Skip to content

Per-tenant rate limits & quotas #4796

Description

@JSONbored

Problem: Without limits, one customer's loop could consume disproportionate compute or starve others, and a runaway loop has no cap.

Area: Platform

Proposal: Apply per-tenant rate limits and hard resource quotas tied to each tenant's paid/staked allocation from #4792.

Deliverables:

  • Enforced per-tenant limits, with a clear, actionable error when a tenant exceeds their quota.

Acceptance criteria:

  • A test tenant exceeding its quota is stopped cleanly, with the correct, user-facing explanation of why.
  • One tenant exceeding its quota has no observable effect on another tenant's loop.

Boundaries:


Part of #4778.


Cross-reference (added after #5669): see #5669 for the reconciliation between this issue and AMS Cloud Readiness's #5223 (Design: per-tenant scheduling and queue-fairness model).

Metadata

Metadata

Assignees

No one assigned

    Labels

    gittensor:featureGittensor-scored feature linked to a feature issue — scores a 0.25x multiplier.

    Projects

    Status
    In Progress

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions