hc-httpclient-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "KARR, DAVID (ATTSI)" <dk0...@att.com>
Subject RE: Is it possible that HttpClient might use more packets under load than not under load?
Date Wed, 28 Jul 2010 19:36:31 GMT
> -----Original Message-----
> From: Oleg Kalnichevski [mailto:olegk@apache.org]
> Sent: Tuesday, July 27, 2010 5:00 AM
> To: HttpClient User Discussion
> Subject: Re: Is it possible that HttpClient might use more packets
> under load than not under load?
> 
> On Mon, 2010-07-26 at 13:13 -0700, KARR, DAVID (ATTSI) wrote:
> > I'm still struggling to understand the infrequent occurrence of
> service
> > calls in our system that take well past the "socket timeout" value
> that
> > we set for the transaction.  I know that the "socket timeout" is PER
> > PACKET, not PER TRANSACTION.
> 
> To be exact it is the maximum _interval_ between two consecutive
> packets. This has nothing to do with the size or number of packets.
> 
> > One thing I'm wondering is whether the sizes of packets could be
> > variable.  For instance, will I end up with smaller packets when the
> > server is heavily loaded, as opposed to larger packet sizes under
> normal
> > load?
> 
> It is perfectly normal that under heavy load intervals between packets
> increase. If, however, those intervals remain below the timeout value
> the connection will not expire despite transactions taking much longer.

That's exactly what I'm looking for, thanks.
Mime
View raw message