tag:blogger.com,1999:blog-3946011063058389308.post1204187433217875726..comments2011-08-24T08:19:47.530-07:00Grid Dynamicshttp://www.blogger.com/profile/18125799569183836823noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-3946011063058389308.post-68494654560562983052011-08-10T14:30:58.757-07:002011-08-10T14:30:58.757-07:00The indexed entryUUID vs indexed cn results are interesting. What was the average length of your cn values? Have you tried a test where the cn values are the same size as entryUUIDs?<br /><br />Note that equality indexing is based on a 4-octet hash of the original value. Possibly there are too many hash collisions, resulting in many entries needing to be checked for a given index lookup. (You could determine if this is happening by watching the debug output from from back-bdb.)<br /><br />Also entryUUID is hashed using its binary form (16 octets), not the text form (36 characters).<br /><br />I had not seen your results until a colleague pointed me to this blog today. Did you ever discuss these results on the OpenLDAP mailing lists? Relying on random chance to get the attention of the OpenLDAP Project isn&#39;t a sound way of getting improvements made...hychttp://www.blogger.com/profile/15473250487285924085noreply@blogger.comtag:blogger.com,1999:blog-3946011063058389308.post-38068120560624561982010-03-09T02:55:40.355-08:002010-03-09T02:55:40.355-08:00I used trivial load generation utility written in C.Dmitry Korotkovhttp://www.blogger.com/profile/07023646962814250784noreply@blogger.comtag:blogger.com,1999:blog-3946011063058389308.post-91674778391172982882009-12-22T08:45:24.524-08:002009-12-22T08:45:24.524-08:00what tool did you use then for the testing?Anonymousnoreply@blogger.com