0

We use server as a an aggregator with lots of small income HTTP requests (~1000 RPS). When we reached that number of RPS, http requests started to take randomly longer amount of time (100ms-4 seconds). This is not an issue with our application server because this delays happen even when we try access statics html page via NGINX (http://5.153.4.91/ok.html).

Is it possible to somehow investigate this issue and determine if this is limitation of network card/network infrastructure/bad kernel settings/misconfigured server?

Ready to show any config files.

1 Answer 1

1

Use scientific method.

Install monitoring.

  • Gather statistics.
  • Make decisions about your pinch points based on the information collected.
  • Make changed based on the decisions made.
  • Monitor the effectiveness of the change.
  • Rinse and repeat.
5
  • What kind of monitoring? Which statistics to gather? Right now your answer could be applied to troubleshooting any performance problem :)
    – Daniil
    Jun 25, 2014 at 12:18
  • 1
    I think that would be rather the point.
    – Sobrique
    Jun 25, 2014 at 12:20
  • That's correct - that's what sysadmins do don't we ? As it stands you know nothing, so start at the top and work down. There is no magic wand to wave that will magically tell you that the plugh xyzzy needs tweaking. Regarding the monitoring, we don't do product recommendations but if we did they would probably be OS dependent and you didn't mention it :(
    – user9517
    Jun 25, 2014 at 12:22
  • Actually I'm not sysadmin, but I have to face this problem. I have standalone hardware server + Ubuntu 12.04 installed. I don't ask for the solution of the problem, but tools/hints that would help to investigate it.
    – Daniil
    Jun 25, 2014 at 12:27
  • 1
    That may be true but you are acting as a sysadmin so get with the programme.
    – user9517
    Jun 25, 2014 at 12:40

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .