CODEWITHSUNDEEP
gogoroutinegorilla
Rakshita Jain
Rakshita Jain

Reputation: 149

Using Go routines in gorilla/mux framework

We are using gorilla mux framework for handling web requests which I suppose automatically runs on all the cpu cores. Is there a benefit of using go routines in such a case for cpu intensive processes for eg looping through a large object?

Upvotes: 3

Views: 2890

Answers (1)

Jonathan Hall
Jonathan Hall

Reputation: 79536

I suppose automatically runs on all the cpu cores.

You suppose wrong. Sort of.

As of Go 1.5, Go will use all of your cores, by running go routines on different cores. But if you don't use go routines, there's no way for it to take advantage of this.

Is there a benefit of using go routines in such a case for cpu intensive processes for eg looping through a large object?

There can be. But you're asking the wrong question.

You don't use Go routines primarily to take advantage of different CPU cores (although this can also be a benefit). You use Go routines to keep your program from blocking while doing something that takes a while.

In the case of a web application, most requests are usually not CPU intensive at all. But they usually spend a lot of time (in computer terms) waiting around for things to happen. They wait for DNS lookups on the request hostname, they wait for the database to look up user credentials to establish a session, they wait for the database to store or return rows to produce the HTTP response, etc.

Without go routines, while doing these things, your server would be unable to do anything else. So if your typical HTTP request took, say, 1 second, to look up DNS, validate an authorization cookie, look up results from a database, and send a response, no other HTTP client could be served simultaneously.

Fortunately, the http package, which is used by Gorilla (and practically every other web framework for Go) already uses Go routines to handle requests. So you're already using (at least) one Go routine per HTTP request.

Whether it makes sense to use additional go routines is more up to your application design than "using more CPU cores."

Some suggestions:

  1. If you're doing multiple things to serve a request that could be done in parallel, use go routines. Suppose you need to do 3 DB queries, you can do each one in a Go routine, so they run simultaneously. They may run on different CPU cores, but that's completely irrelevant to actual performance, since each one is essentially doing nothing, and just waiting for the database. But you'll still improve performance. (For completeness sake,I'll mention that goroutines aren't the only way to run database queries in parallel--but they are the idiomatic Go way, and the easy way in Go).
  2. If you have tasks that are run by an HTTP request, but don't affect the HTTP response, you can run them in a Go routine, so your response can return sooner. Logging is a common example. You probably want to log your HTTP requests, but if the logger is slow, you probably don't want to wait for logging to finish before sending your HTTP response. So do it in a go routine. The logger can take it's sweet time and finish after the client has already received their response. Another example is sending emails--please don't wait for an email to be sent before responding to an HTTP client.

Upvotes: 11

Related Questions