I just wanted to share my experiences using taskspooler (ts) for server side compute jobs.
In a web application, HTTP requests should return quickly meaning that lengthy calculations should be performed asynchronously.
I have used taskspooler a couple of times for machine learning jobs to manage a queue
The alternative is something like AWS Simple Queue Service (SQS), but to be honest if you're not doing auto-scaling to meet a constant time target, taskspooler is perfectly adequate.
Typically, I get the web-application (any language, Python or e.g. Rails) to issue a system call which runs a script to enqueue a job with taskspooler. The job itself is then run by taskspooler and then does a local HTTP request to PUT or POST the results back to the web app, also acting as a notification that the job has finished.
Anyway I thought someone might find this a useful way to run small machine learning jobs in a queue as part of a webapp.