Platform engineering, system design and infrastructure shape.
7 items at advanced level · all topics
When would you scale vertically instead of horizontally?
Go vertical when the bottleneck doesn't split: database buffer cache, cache memory, network bandwidth on one node, a single-threaded lock. Go horizontal when the work partitions cleanly. The real question isn't which is better, it's whether the thing that's saturated can be divided across machines at all.
What happens if an Auto Scaling instance never becomes healthy?
Auto Scaling terminates and replaces an instance that never passes health checks, which is fine once and expensive in a loop. How long you burn depends on the grace period, the check thresholds, and which health check type the group is actually using. The fix is working out why it's unhealthy: usually startup config, a wrong health endpoint, or a grace period shorter than boot time.
Your 5 GB monorepo has code, binaries, Terraform and Helm charts, and clones are slow. What do you do?
Measure what the 5 GB actually is: big blobs, deep history, and a wide tree are three different problems with three different fixes. Work through them in order of cost, artifacts out of Git, binaries into LFS, partial clone and sparse checkout for developers, and treat splitting the repo as the last resort rather than the first idea.
Your organization wants production deployments to be reproducible six months later. How does Git help?
Git pins the source exactly and that's all it pins. Real reproducibility also needs locked dependencies, versioned infrastructure, immutable artifacts, and a stored record linking a deployment to its digest, plus keeping the actual artifact, because rebuilding it later is rarely byte-identical.
A company uses AWS Organizations with all features enabled. For compliance reasons, workloads may only run in two approved AWS Regions. Account administrators in member accounts currently hold the AdministratorAccess policy. The company must prevent resources from being created in any other Region. What should a solutions architect do?
A service control policy is the only control that an account administrator cannot remove, because it caps what identities in a member account may do no matter what their IAM policies say. IAM policies and permissions boundaries can both be edited by the very administrators you are trying to constrain.
A company runs a business-critical application in one AWS Region. The disaster recovery plan requires a recovery time objective of 10 minutes and a recovery point objective of 1 minute in a second Region. The company wants to avoid paying for a full duplicate of the production fleet. Which disaster recovery strategy should a solutions architect choose?
RTO and RPO in minutes rule out backup and restore and pilot light, because both require standing infrastructure up before traffic can move. Warm standby keeps a scaled-down copy always running, which is what buys minutes instead of hours without paying for a second full fleet.
A multiplayer game server runs on Amazon EC2 instances behind Network Load Balancers in three AWS Regions. Players connect over UDP. The company needs to route each player to the lowest-latency healthy Region, to fail over within seconds if a Region becomes unhealthy, and to publish a fixed set of IP addresses that players' firewalls can allow. Which solution meets these requirements?
UDP, static IP addresses and fast regional failover all point at AWS Global Accelerator rather than CloudFront. Global Accelerator gives you anycast static IPs, carries traffic over the AWS backbone, and reroutes without waiting for DNS to expire anywhere.