Separate jobs with very different runtimes. A quick report shouldn't spend hours behind a huge export. Separate queues let you allocate workers and tune each workload independently. More on backpressure and async workers:
Scale workers with the backlog. Queue depth tells you how much work is waiting. Add capacity as it grows, while checking that downstream services can handle the extra concurrency.
A queue won't save you if jobs arrive faster than workers can finish them. It just turns overload into a longer wait. Three ways to keep that under control.
The best system design material is written by the people running the systems. Five worth reading, starting with Netflix on chaos engineering and streaming to 230 million people. netflixtechblog.com
Which changes what is being scored. Not whether you can produce the function, but whether you can direct the tool, catch it when it is wrong, and own the result.
Meta has a coding round where you keep the AI. It is trying to hold two things at once: the practical reality that you will use a model at work, and the need to evaluate the person behind the keyboard.
When you meet someone, ask what you should know that isn't obvious and who else you should talk to. You'll learn things the onboarding doc misses. More on building context, relationships and momentum in a new role:
Take on a small, useful change you can finish well. Learn the code, get it reviewed and follow it through to production. Your teammates get evidence that you can deliver, and you learn how work actually gets done there.
Persistence stays off the critical path. AOF appends sequentially with configurable fsync, RDB snapshots run in a child process. The tradeoff is you can lose data, and Redis makes that trade to stay fast.
Redis is single-threaded on purpose. One core loop processing commands means no locks, no context-switch thrash, no coordinating threads over shared data. Want more CPU? Run more instances in a cluster.
This one shows up in infrastructure-flavoured interviews specifically, where the interviewer wants to see how you reason about indexing rather than which product you would buy.
Design post search, and the interviewer tells you Elasticsearch is off the table. Now you have to build the inverted index yourself, which is the whole point of the question.
52 Followers 139 FollowingEngineering - Amazon | Building software by day, dreams by night 💻 | Builder by Passion | Mentoring devs | Football fanatic | Movie buff | Meme connoisseur
1K Followers 914 FollowingOlder than all dogs. Creating: Hello Interview. Sold 2 startups, failed 1, ready for more. Previously ML eng leader FB, AMZN.