Bernd Rinn <bernd@sdf.lonestar.org> wrote on 09/14/2005 02:42:22 PM:
> We are experiencing very bad performance when performing an update
> triggered from the synchronize view. When performed from the Java
> perspective the update takes 23 s, when performed from the synchronize
> view it takes > 1300s (I don't know how much longer since I went to
> lunch and when I came back an hour later it was finished).
>
> We are using 0.9.34. I tried JavaHL and JavaSVN, but it doesn't seem to
> make a difference (note: it's not about the synchronize procedure itself
> but about an update triggered from synchronize).
>
> The strange thing is that I can't spot a bottleneck: the client CPU is
> idle, the network connection is idle, the server CPU is idle, and so is
> the server disk. Everything seems to wait on something but I don't know
> what. Even stranger, the operation is faster when performed on a DSL
> line than when performed on a 100MBit/s network.
>
> Has anyone seen a similar phenomenon? Any ideas what the reason could
be?
Have you used the Synch view in previous releases? I am trying to decide
if this is a new problem or not. We are aware there are some performance
issues with that action, particularly with JavaSVN. At the same time, we
also had to make some important correctness fixes in this area in 0.9.34
so I am trying to ascertain whether that also made performance worse. In
my own testing, it was equal or even better than it was in 0.9.33 and .32.
The fact that you are seeing no activity makes it seem like it is
something else like a deadlock of some sort. There are a lot of internal
cache updates that have to occur after an update, statuses have to be
rechecked, builds performed etc...
Mark
_____________________________________________________________________________
Scanned for SoftLanding Systems, Inc. and SoftLanding Europe Plc by IBM Email Security Management Services powered by MessageLabs.
_____________________________________________________________________________
Received on Thu Sep 15 06:52:23 2005