On 08.06.2012 20:58, Philip Martin wrote: > "C. Michael Pilato" <cmpil...@collab.net> writes: > >> On 06/08/2012 01:41 PM, Mark Phippard wrote: >>> On Fri, Jun 8, 2012 at 12:01 PM, <cmpil...@apache.org> wrote: >>>> Author: cmpilato >>>> Date: Fri Jun 8 16:01:00 2012 >>>> New Revision: 1348131 >>>> >>>> URL: http://svn.apache.org/viewvc?rev=1348131&view=rev >>>> Log: >>>> Finish issue #4179 by eliminating no-op HEAD requests from ra_serf's >>>> checkout/update logic. >>> Nice! This makes your other change of eliminating GET requests really >>> fly, at least in an idealized scenario. >> Thanks for providing some numbers. I'm on a flaky network today. I can't >> even complete two consecutive updates from the ASF repos. :-( > Figures from mod_status on my local mirror for the second update below: > > svn co http://localhost:8080/repos/asf/subversion/trunk wc > svn up --set-depth empty wc > svn up --set-depth infinity wc > > For a 1.7 client the server reports: > > Total accesses: 4133 - Total Traffic: 45.9 MB > CPU Usage: u2.7 s.9 cu0 cs0 - 20% CPU load > 230 requests/sec - 2.6 MB/second - 11.4 kB/request > > For a 1.8 client the server reports: > > Total accesses: 2188 - Total Traffic: 3.3 MB > CPU Usage: u1.64 s.58 cu0 cs0 - 22.2% CPU load > 219 requests/sec - 341.8 kB/second - 1599 B/request
That's ... amazing. I wonder where we'll pay for it. -- Brane
