On Mon, May 19, 2008 at 7:01 AM, Martin Langhoff <[EMAIL PROTECTED]> wrote: > Working on ds-backup, I am concerned about performance of extracting > all of the datastore metadata. Both Tomeu and Ivan have warned me > about performance and memory impact. I want to repro the problem asap. > > If a full metadata dump can be acceptably fast, a lot of complications > and error conditions in the ds-backup code just disappear, thanks to > rsync. I think it should be fast - our storage space limits the amount > of data we are dealing with in a normal case, and the machine is fast > enough that we should be able to burn through a few thousand entries > in less that 2s. But I need a good test case. > > (So far, with small datasets I haven't been able to go over 1s, and > the python startup dominates time anyway). > > Has anyone got a good sized datastore that I can grab? I am happy to > respect privacy of any personal datastore sample... but you must cope > with my browsing a bit around. :-) My test machines are on 703 - happy > to up/down/side-grade in case that matters.
I never had a _real_ DS with the thousands of entries that a kid can create in not a very long time, but I can give a you a script that creates something more or less close to it. We could take a DS with dozens or a few hundreds of real entries, and multiplicate those until you have the desired size. Do you think this would be enough? Regards, Tomeu Do you think it would be enough? _______________________________________________ Devel mailing list [email protected] http://lists.laptop.org/listinfo/devel
