praxstack Backend Principle Eng C++ Pro Max
Principal-level guidance for C++ backend systems, low-latency services, and infrastructure. Emphasizes correctness, memory safety, and predictable performance.
When to Apply
- Designing or refactoring C++ backend services and infrastructure
- Reviewing code for memory safety, concurrency, and latency regressions
- Building high-throughput networking, storage, or compute systems
- Incident response and performance regressions
Priority Model (highest to lowest)
| Priority | Category | Goal | Signals |
|---|---|---|---|
| 1 | Correctness & UB Avoidance | No undefined behavior | RAII, invariants, validated inputs |
| 2 | Reliability & Resilience | Fail safe under load | Timeouts, backpressure, graceful shutdown |
| 3 | Security | Hard to exploit | Hardened builds, safe parsing, least privilege |
| 4 | Performance & Latency | Predictable P99 | Stable allocs, bounded queues, zero-copy where safe |
| 5 | Observability & Operability | Fast triage | Trace ids, structured logs, metrics |
| 6 | Scalability & Evolution | Safe growth | Statelessness, sharding, protocol versioning |
| 7 | Tooling & Testing | Sustainable velocity | Sanitizers, fuzzing, CI gates |
Quick Reference (Rules)
1. Correctness & UB Avoidance (CRITICAL)
raii- Own resources with RAII and deterministic lifetimesno-raw-ownership- Raw pointers only for non-owning referencesbounds- Validate all indices and sizes at boundariesinvariants- Assert core invariants and state transitionstime- Use monotonic clocks for durations
2. Reliability & Resilience (CRITICAL)
timeouts- Explicit timeouts for every external callbackpressure- Bounded queues; apply load sheddingshutdown- Drain in-flight work with deadlinesbulkheads- Isolate thread pools by dependency
3. Security (CRITICAL)
safe-parse- Validate untrusted input; avoid unsafe string opsharden- Compile with stack protection, PIE, RELRO, FORTIFYsecrets- No secrets in logs or core dumpsleast-priv- Drop privileges and sandbox when possible
4. Performance & Latency (HIGH)
allocs- Minimize allocations in hot pathscopy- Prefer move or views; avoid unnecessary copiescache- Improve locality; avoid false sharingio- Use async I/O where appropriateprofiling- Measure before optimizing
5. Observability & Operability (HIGH)
logs- Structured logs with request and trace idsmetrics- RED/USE plus business KPIstracing- Propagate trace context across threadscrash- Symbolized crash reports and core dump policies
6. Scalability & Evolution (MEDIUM)
stateless- Externalize state, enable horizontal scalepartitioning- Shard by stable keysversioning- Protocol and schema versioninglimits- Explicit limits on payloads and queue sizes
7. Tooling & Testing (MEDIUM)
sanitizers- ASan, UBSan, TSan in CIfuzzing- Fuzz parsers and protocol handlerstests- Unit, integration, and load testslint- clang-tidy, clang-format, warnings as errors
Execution Workflow
- Clarify latency/SLOs, throughput, and cost budgets
- Map data flow, thread model, and failure modes
- Define interfaces and memory ownership contracts
- Implement with bounded queues and explicit timeouts
- Add observability and crash diagnostics
- Validate with sanitizers, fuzzing, load tests
- Review risks and publish runbooks
Language-Specific Guidance
See references/cpp-core.md for toolchain defaults, concurrency patterns, and hardening.