Asynchronism

reduce request times for expensive operations by promising to do them later, so that the actual request can finish.
can also be used to precompute values ahead of time, such as encoding videos or generating thumbnails.
message-queue.
- a queue to receive, hold, and deliver messages.
- application pushes a job request to the queue, and will notify the user of job status and updates.
- a worker picks up the job from the queue, completes the request, and signals to the user that the job is complete.
- while the job is enqueued, we can show the end user a temporary local version until the job is actually finished.
- popular tools:
  - redis can be used as a simple message-queue but messages may be lost.
  - rabbitmq.
  - amazon-sqs is popular but has high latency.
task-queue.
- receive tasks and related data, run them, and then deliver the results.
- very similar to the message-queue, but the tasks on a task queue specify what needs to be done, whereas a message-queue simply stores messages that workers will read.
- popular tools:
  - celery.
back-pressure.
- when queues grow to be larger than memory, requests start getting dropped because they can’t be stored in memory.
- we tell users that a server is busy to create pressure for these requests to slow down, so our queue can catch up.
disadvantages.
- adds complexity.
- when it’s just simple usecases, or when there’s not enough requests, synchronous operations without queues might just make more sense.

table without id file.inlinks as Backlinks
where file.name = this.file.name

References.

Categories:: scalability