coder-system-design-perf-patterns

name	coder-system-design-perf-patterns
description	System-level performance pattern selection with measurable tradeoffs and SLO alignment.

<when_to_use> Designing architecture under latency, throughput, or cost pressure Planning scaling strategy for growth or traffic spikes Reviewing performance bottlenecks and reliability tradeoffs </when_to_use>

<input_requirements> Current bottleneck description and baseline metrics Target SLO/SLI and error budget constraints Traffic profile (steady, bursty, hot keys, read/write ratio) Operational constraints (cost, team capacity, infra limits) </input_requirements>

<pattern_catalog> Cache-aside for read-heavy paths with tolerated staleness Queue-based load leveling for burst absorption Async processing for long-running tasks and user-facing responsiveness Batching to reduce per-operation overhead and network chatter Load shedding to protect core goodput under overload Horizontal scaling with autoscaling tied to meaningful saturation metrics Hotspot mitigation via key design and partition distribution </pattern_catalog>

<selection_rules> Choose patterns from measured bottlenecks, not preference Define expected gain and rollback trigger before rollout Account for consistency, cost, and ops complexity tradeoffs Keep observability and SLO guardrails in same design package </selection_rules>

<tradeoff_matrix> Caching: lower latency, possible staleness and invalidation complexity Queues and async: smooth peaks, eventual consistency and replay complexity Batching: higher throughput, partial failure handling complexity Load shedding: preserves core capacity, intentional selective failures Horizontal scaling: capacity gain, bottleneck may shift to shared dependency </tradeoff_matrix>

<do_not> Do not scale on vanity metrics unrelated to user SLI Do not use unbounded retries/queues under overload Do not evaluate performance on averages only; include tail latency Do not ship performance changes without capacity and rollback plan </do_not>

<output_requirements> Chosen pattern(s) with why-now rationale Expected SLI impact and acceptance thresholds Operational tradeoffs and failure modes Rollout, monitoring, and rollback plan </output_requirements>

AWS Builders Library: Load Shedding Google SRE: Service Level Objectives Google SRE Workbook: Alerting on SLOs Google SRE: Cascading Failures Azure Queue-Based Load Leveling Kubernetes HPA DynamoDB Partition Key Design