← All patterns The problem An upload burst creates more processing work than consumers can finish immediately.
When it fits Work can complete asynchronously and the product can communicate acceptance separately from completion.
The whole architecture
Architecture diagram Download SVG An acceptance path buffers jobs while a competing worker pool drains retained work.
Blue arrows are calls or work acquisition; purple arrows show asynchronous job delivery.
queue load leveling architecture An acceptance path buffers jobs while a competing worker pool drains retained work. Blue arrows are calls or work acquisition; purple arrows show asynchronous job delivery. Accept work Competing worker pool clients → api: submit api → queue: enqueue queue → worker-a: job queue → worker-b: job worker-a → output: effect worker-b → output: effect submit enqueue job job effect effect Acceptance and completion are separate outcomes. Producers Submit asynchronous work Report acceptance only after the chosen durable enqueue contract succeeds. Ingress API Validate and enqueue A job is assigned to a consumer; this is not independent-subscriber fan-out. Work queue Retained backlog Acknowledge after the required effect; a crash can cause redelivery. Worker A Competing consumer This worker shares the same subscription or work queue. Worker B Competing consumer An atomic local effect and job-identity record can protect against duplicate processing. Output SQL Effect + stable job ID Sustained arrivals above worker capacity still require backpressure or more processing capacity. Backlog controls Age, bounded retries, dead-letter handling API / data call Event / job flowSelect a component, or focus it and press Enter, to read its role.
Each worker competes for work; adding workers does not create another independent subscription. Buffering absorbs temporary bursts but consumes time and storage. Walk through the design
Follow one operation Core flow Failure & recovery
Select a step to highlight its participants. This conceptual walkthrough is separate from challenge scoring.
Arrows show the conceptual relationships used by this walkthrough.
1 Accept asynchronous work2 Buffer the burst3 Complete and acknowledgeStep 1 of 3 Accept asynchronous work App validates the request and publishes a durable job before reporting acceptance.
← Previous Restart Next step →
Trade-offs to make explicit Buffering absorbs a temporary burst at the cost of delay and queue storage. A queue cannot fix sustained arrivals above processing capacity. Size consumers and set backpressure limits. Failure modes to plan for A consumer can crash after an effect but before acknowledging; redelivery needs idempotent handling. A poison message needs bounded retries and a recovery path rather than blocking useful work. Before you build Explain your decision How would you set a backlog-age limit and respond when arrivals exceed consumers for an hour?
Use Plan & estimate inside practice to record your assumptions and trade-offs.
These links open the original challenges. Existing round unlocks, completions and points apply.
Upload once, publish when ready Practice queued processing, connected workers and completion deadlines.
Open challenge → One provider stalled; the others still deliver Practice independent provider recovery and healthy-channel latency.
Open challenge → Further reading Microsoft · Queue-based load leveling ↗
Original PractiseGrid explanations and conceptual diagrams, with primary references for further study. Dedicated scored exercises can be added as coverage grows.