/aws
S3, IAM, Lambda, EC2 — how each service actually behaves in production.
Notes on the AWS pieces that show up in day-to-day backend work: S3, SQS, SNS, Lambda, DynamoDB, CloudFront, and the glue between them. Everything here comes from things I've built and shipped, not from certification slides. Expect small, focused walkthroughs and honest failure modes.
Buckets, keys, presigned URLs, event notifications, storage classes.
1 article → IAMPolicy evaluation, roles, and a recipe for scoping credentials so a leak stays boring.
1 article → LambdaCold starts, execution model, VPC networking, and when Lambda is the wrong answer.
1 article → SQSPer-consumer queues, visibility timeouts, DLQs, and idempotent workers.
2 articles → SNSTopic semantics, filter policies, and fan-out patterns for event-driven backends.
2 articles → DynamoDBSingle-table design, GSIs, capacity modes, and the queries that decide your schema.
Coming soon →Retries are guaranteed in distributed systems. Duplicate side effects are not. A deep dive into idempotency keys, the claim-first ledger pattern, and how to implement it in Django and PostgreSQL.
Synchronous, asynchronous, and poll-based invocation give the same Lambda code completely different failure modes, retry behavior, and scaling. A field guide to choosing, plus the three front doors for HTTPS: Function URLs, API Gateway, and ALB.
Build the SNS-to-SQS fan-out pattern where each consumer owns a private queue with its own retry policy and DLQ. Covers queue policies, raw message delivery, redrive config, and the delivery guarantees you actually get.
A 25-minute walkthrough: create an AWS account, lock away the root user, create an IAM admin daily driver with MFA, and set billing budgets so you never get a surprise bill.
Upload files from the browser straight to S3 — no proxying through your backend. Presigned PUT URLs with Django, the s3v4 SignatureDoesNotMatch fix, and the exact CORS config you need.
Both make your system asynchronous, but they solve different problems. Queue semantics vs topic semantics, when to reach for each, and the SNS → SQS fan-out pattern that shows up in real system designs.
Why AWS won the cloud, what it looks like from a backend engineer's seat, and the one setup mistake almost every beginner makes on day one.
S3 event notifications → SNS fan-out → SQS-backed workers that parse, classify, and index uploaded documents. Idempotent by design.
Django issues presigned PUT URLs; a Lambda triggered on ObjectCreated runs ClamAV, tags clean objects, and quarantines the rest.
CloudWatch Logs → Kinesis Firehose → S3 in a dedicated audit account, with lifecycle policies moving cold logs to Glacier.
Real AWS integrations — S3 presigned uploads, event-driven pipelines — built on Django.