On Tue, Mar 02, 2004 at 07:06:31PM +0100, Dieter Maurer wrote: (snip)
I interpreted these observations in the following way:
* time is dominated by the disc access
* time is largely independent of the object size (if the object is in the OS disc cache)
* ZEO overhead is about 1 to 3 ms per object.
Hmm. I wondered if ZEO overhead per-object was significant, but this seems not too bad. If I read OFS/Image.py correctly, the pdata chunks are 65536 bytes. So a 40 MB file would split up into about 640 chunks, so ZEO overhead by your calculations should be roughly 1-2 seconds. This does not account for the order of magnitude difference I am seeing (initial load from filestorage: 0:03.87, initial load from zeo: 0:33.54). -- Paul Winkler http://www.slinkp.com Look! Up in the sky! It's PROTO FLASH TROMBONIST! (random hero from isometric.spaceninja.com)