mungohill has asked for the wisdom of the Perl Monks concerning the following question:
Esteemed Monks,
I am using WWW:Curl::Easy to send a very small update message. Because of the way the service works, I need to do a load of individual performs, in some cases running into the hundreds to send all the required updates. A user is sitting staring at a screen while this is happening and waiting for a result.
Running my client program on, let's call it Machine A, I'm seeing 'verbose' output like this:
* About to connect() to blah.blah.com port 8080 * Trying xxx.xxx.xxx.xxx... * connected * Connected to blah.blah.com (xxx.xxx.xxx.xxx) port 8080 > POST /submit.sp HTTP/1.1 Host: blah.blah.com:8080 Accept: */* Content-Length: 5 Content-Type: application/x-www-form-urlencoded blah < HTTP/1.0 200 ITAS Ltd SOAP Server v1.3.53 < Content-length: 343 < Server: ITAS Ltd SOAP Server v1.3.53 < Connection: close * Closing connection #0
This works, well, not exactly quickly, but easily fast enough for my purposes. On, let's call it Machine B, I see this:
* About to connect() to blah.blah.com port 8080 (#0) * Trying xxx.xxx.xxx.xxx... * connected * Connected to blah.blah.com (xxx.xxx.xxx.xxx) port 8080 (#0) > POST /submit.sp HTTP/1.1 Host: blah.blah.com:8080 Accept: */* Content-Length: 1523 Content-Type: application/x-www-form-urlencoded Expect: 100-continue * Done waiting for 100-continue * HTTP 1.0, assume close after body < HTTP/1.0 200 ITAS Ltd SOAP Server v1.3.53 < Content-length: 277 < Server: ITAS Ltd SOAP Server v1.3.53 < Connection: close < * Closing connection #0
On Machine B it takes about a second to execute. You'll notice that there was 'Expect: 100-contine' which wasn't present on Machine A, and indeed, the delay on Machine B is between 'Expect: 100-continue' and '* Done waiting for 100-continue'.
To me, this suggests that, on Machine B, the socket has been created with an SO_KEEPALIVE option. Presumably in response to some default socket options. I know that the keepalive is set up differently on the two machines. Frankly the updates I am doing are not the most crucial things in the world. It is far more important to me that they are done quickly rather than 100% reliably, so I'm looking at ways of speeding it up.
Do you think I'm right in my ramblings about keepalives?
If so, do you think there's any way that I can force curl to create it's socket without keepalives?
|
|---|
| Replies are listed 'Best First'. | |
|---|---|
|
Re: Curling slowly
by Anonymous Monk on Mar 29, 2012 at 10:33 UTC | |
by mungohill (Acolyte) on Mar 29, 2012 at 10:46 UTC |