This website uses cookies
Read our Privacy policy and Terms of use for more information.
Software Letters
Sharing insights about Software engineering, Design patterns, AI applications and NoCode tools.
Written by
Houssem BEN SLAMA
Connect
SQS
+7
Oct 15, 2025
•
4 min read
Tracing async workflows, debugging across queues, and the metrics that actually matter—an advanced guide for serverless teams.
AWS
+3
Aug 3, 2025
7 min read
AWS Glue is a fully managed ETL (Extract, Transform, Load) service designed to simplify data integration. It automates much of the effort in building, maintaining, and running ETL jobs by providing a serverless environment, built-in transformation logic, and native integration with other AWS services like S3, DynamoDB, and OpenSearch.
DDD
+4
Oct 3, 2024
12 min read
Adapting the C4 Model for Serverless, Domain-Driven Design (DDD), and Event-Driven Architecture (EDA)
Data-lake
+8
Jul 2, 2024
9 min read
Understanding Apache Kafka’s Architecture, Benefits, and How It Outperforms Other Data Platforms
Kubernetes
Jun 5, 2024
10 min read
From Basic Operations to Advanced Techniques
Containerization
Jun 4, 2024
Unleashing the Power of Containerization and Orchestration for Modern Applications
GitHub
+1
Nov 18, 2026
6 min read
GitHub split its busiest MySQL cluster without a rewrite by drawing domain boundaries in the app first, enforcing them with linters, then moving 130 tables in a cutover of a few tens of milliseconds. The three mechanisms, their costs, and how to borrow them at your size.
Tutorial
+2
14 min read
Amazon MSK is Apache Kafka with the brokers (or the whole cluster) run by AWS. The model, the Serverless vs Express vs Standard decision with real prices, the limits that bite, and a build: a Serverless cluster, a keyed producer, a two-member consumer group you can watch rebalance, and a lag check.
Nov 17, 2026
Step Functions is a durable state machine you rent by the transition or by the second. Here is the model, the Standard vs Express decision with real numbers, the limits that bite, and a saga built with zero Lambda functions that compensates correctly even when it cannot tell whether a step happened.
Canva
Canva collects 25 billion product events a day with a pipeline whose most important design decision is not the stream. It is the rule that every event schema must stay readable forever, plus a cheap pipe with a fallback that hides its tail latency.
design_patterns
Nov 16, 2026
The Composite pattern is one interface shared by leaves and containers. The version that survives production is that interface plus four rules: one parent per node, no cycles, a traversal you control, and a policy for cached aggregates. Here is each rule, with code, and where the DOM and AWS CDK enforce it.
Dropbox
Dropbox spent a year and a half pushing its Python monolith toward team-owned services, then stopped. Atlas keeps the code close to a monolith, gives every group of routes its own cluster, and hands all the operations to a platform team.
11 min read
EventBridge relaunched its event bus in September 2026. The old one is now "classic", the new one bills per GB and replays from retention. Here is how to pick between them, Pipes, and Scheduler, and a build that proves filtering, dedup, a DLQ, and replay.
Cloudflare
Nov 10, 2026
When Cloudflare replaced NGINX with Pingora, the 80 ms it cut from p95 latency came from changing who owns the connection pool, not from faster code. Here is how a process model decides your latency, and what to copy from it.
SNS
SNS is the service that turns one publish into many deliveries, and then quietly discards the ones that fail. We build a topic that fans out to three queues with filter policies, add a dead-letter queue, and a FIFO topic with replay. Plan for about 45 minutes and a few cents.