7 min read
Designing p99 latency budgets for API gateways
Why averages hide the latency your users feel, how to split a tail-latency budget across gateway stages, and how to verify it under load.
API
Research
Long-form analysis of how high-throughput server-side systems are designed, scaled, secured and kept running. Regulation appears only where it shapes engineering decisions.
7 min read
Why averages hide the latency your users feel, how to split a tail-latency budget across gateway stages, and how to verify it under load.
API
7 min read
How to make POST requests safe to retry with idempotency keys: storage design, request fingerprinting, concurrency and expiry.
API
6 min read
Why writing to a database and publishing to a broker in the same request goes wrong, and how the outbox pattern, change data capture and idempotent consumers fix it.
Architecture
6 min read
A practical comparison of rate-limiting algorithms, how to enforce limits consistently across gateway replicas, and how to communicate limits to clients.
Scaling
7 min read
How mutual TLS authenticates services to each other, how to issue and rotate short-lived workload certificates, and how to turn identity into authorisation.
Security
7 min read
Why bearer tokens are risky for high-value APIs, how DPoP and mTLS-bound tokens tie a token to a key, and what the OAuth 2.0 Security Best Current Practice recommends.
Identity
6 min read
When to encrypt, when to tokenise and when to use format-preserving encryption – and how to design a token vault and key management that survive an audit.
Security
7 min read
Streaming parsers, XXE hardening, multi-version schema validation, canonical models and quarantine design for high-volume XML message pipelines.
Data
6 min read
How recovery point and recovery time objectives drive replication, topology and runbook choices – and why only rehearsed failover counts.
Resilience
6 min read
How the Digital Operational Resilience Act translates into concrete engineering work: asset inventories, change control, logging, incident classification, resilience testing and third-party exit plans.
Resilience
6 min read
Requests, limits, CPU throttling and the right scaling signals: how to keep p99 latency stable while pods scale up and down.
Scaling
7 min read
A step-by-step approach to replacing a legacy monolith without a big-bang rewrite: characterisation tests, a routing facade, extraction order and data migration.
Architecture