[chandler-users] osaf.us updated to Cosmo 0.6.0.1

[Andre Mueninghoff]

Hi Jared,

Returning from errands I observed the sync of the fourteen collections
to be incomplete after two hours. I left again before it completed. (Am
still out.) Seems like a bizzare fringe case. I'm measuring "incomplete"
by having yet to see a new sync item created in the Sharing Activity
collection, and by seeing the command window continue to scroll through
sharing "(nnn)"-type messages, and by status bar messages indicating
collections sync one by one...the usual stuff. No errors, no tracebacks,
pyton mem usage steady abround 420MB, and python CPU usage at 98-99
basically continously.

I checked the bandwidth meter on cnet.com and got 1,700M bps. (I may be
recalling the units incorrectly, but it was higher than typical cable
modem speeds according to cnet.) 

I haven't had a chance to revert to a previous build to see if a new
check-in might be contributing. Will let you know. 

Andre

On Sat, 10 Mar 2007 12:14:29 -0800, "Jared Rhine" <jared at wordzoo.com>
said:
> Andre Mueninghoff wrote:
> > Is anyone else noticing extremely long durations to sync shares
> > between Chandler r13453 (WinXP) and osaf.us?
> 
> I did a spot check of the server-recorded times for the initial 
> PROPFINDs of the OSAF office calendar and one of your collections, 
> Andre.  This was checked on the 8th (before update) and today (after 
> update).  The initial PROPFIND during a DAV-based sync operation 
> dominates the synchronization time.
> 
> Initial impressions are that there's pretty wide variance in sync times 
> (from about 2 seconds to 11 seconds for the office calendar).  It 
> *might* be higher after the update, if I squint, but now I want to write 
> a quick statistical analysis to put some numbers behind the question.
> 
> If by "extremely long durations", you're implying 2x or more slowly than 
> previously, the server-recorded numbers don't show anything like that. 
> It's certainly possible you're seeing network effects, but that's 
> notoriously difficult to establish.
> 
> I'd like to hear any other observations about sync times.  I sync in the 
> background, so I never would have noticed if the time have actually gone 
> up.  There's other operations than just the initial PROPFIND reported 
> above, so there could be additional effects I haven't looked at.
> 
> Thanks for the report, Andre.  I'll be heightened to the possibility and 
> take an additional look at possible causes and some real statistics.
> 
> -- Jared
> 
> _______________________________________________
> chandler-users mailing list
> chandler-users at osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/chandler-users