[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
RE: Questions...
> -----Original Message-----
> From: owner-openldap-devel@OpenLDAP.org
> [mailto:owner-openldap-devel@OpenLDAP.org]On Behalf Of Howard Chu
> Sent: Monday, January 26, 2004 5:00 AM
These numbers are comparable to the numbers from this paper:
http://www.terena.nl/conferences/tnc2003/programme/papers/p1d1.pdf
> REL_ENG_2 ldbm (2.0.27 + idl.c fixes)
> 879.130u 565.180s 48:35.92 49.5% 0+0k 0+0io 105330pf+0w
> total 332611
> -rw------- 1 hyc users 187510784 Jan 26 02:09 dn2id.dbb
> -rw------- 1 hyc users 151740416 Jan 26 02:08 id2entry.dbb
> -rw------- 1 hyc users 8192 Jan 26 02:07 nextid.dbb
However, the above load was done using the default dbcachesize. Setting the
cache to 384MB brought the time down to
505.150u 102.130s 11:40.75 86.6% 0+0k 0+0io 757pf+0w
This tells me that the dbcachesize was not configured in the tests performed
in the above paper. (dbnosync was also set) (384MB is apparently larger than
the entire database content in this case.)
> REL_ENG_2_1 ldbm
> 801.210u 459.460s 29:13.49 71.8% 0+0k 0+0io 69899pf+0w
> total 587342
> -rw------- 1 hyc users 187617280 Jan 26 02:47 dn2id.dbb
> -rw------- 1 hyc users 411459584 Jan 26 02:46 id2entry.dbb
> -rw------- 1 hyc users 8192 Jan 26 02:45 nextid.dbb
Same here - dbcachesize set to 384MB:
465.160u 108.820s 11:26.04 83.6% 0+0k 0+0io 24563pf+0w
> REL_ENG_2_2_5 ldbm
> 842.710u 463.140s 29:23.94 74.0% 0+0k 0+0io 70607pf+0w
> total 587342
> -rw------- 1 hyc users 187617280 Jan 26 03:21 dn2id.dbb
> -rw------- 1 hyc users 411459584 Jan 26 03:19 id2entry.dbb
> -rw------- 1 hyc users 8192 Jan 26 03:18 nextid.dbb
Oddly enough, the database volume here is identical to in 2.1, but the
performance is slightly off.
496.800u 108.830s 12:01.11 83.9% 0+0k 0+0io 24610pf+0w
Note that the BDB cache size was already set to 384MB in the BDB and HDB
numbers below.
In BDB 4.2 there's a new option to turn off transaction logging. This can
provide another 25% gain for bdb/hdb.
Note that all of these results are for slapadd with no attribute indexing,
which is a bit pointless since searching doesn't really work without an
objectClass index. I still have some trials to run with indexing enabled...
> REL_ENG_2_1 bdb
> 689.950u 56.070s 17:20.57 71.6% 0+0k 0+0io 276994pf+0w
> total 1111581
> -rw-r--r-- 2 hyc users 98 Jan 26 03:31 DB_CONFIG
> -rw------- 1 hyc users 8192 Jan 26 03:49 __db.001
> -rw------- 1 hyc users 503324672 Jan 26 03:49 __db.002
> -rw------- 1 hyc users 2162688 Jan 26 03:49 __db.003
> -rw------- 1 hyc users 368640 Jan 26 03:49 __db.004
> -rw------- 1 hyc users 16384 Jan 26 03:49 __db.005
> -rw------- 1 hyc users 196280320 Jan 26 03:48 dn2id.bdb
> -rw------- 1 hyc users 431816704 Jan 26 03:49 id2entry.bdb
> REL_ENG_2_2_5 bdb
> 714.440u 58.610s 18:19.99 70.2% 0+0k 0+0io 280852pf+0w
> total 1198287
> -rw-r--r-- 2 hyc users 98 Jan 26 03:31 DB_CONFIG
> -rw------- 1 hyc users 8192 Jan 26 04:10 __db.001
> -rw------- 1 hyc users 503324672 Jan 26 04:10 __db.002
> -rw------- 1 hyc users 2162688 Jan 26 04:10 __db.003
> -rw------- 1 hyc users 368640 Jan 26 04:10 __db.004
> -rw------- 1 hyc users 16384 Jan 26 04:10 __db.005
> -rw------- 1 hyc users 196280320 Jan 26 04:09 dn2id.bdb
> -rw------- 1 hyc users 520257536 Jan 26 04:10 id2entry.bdb
> REL_ENG_2_2_5 hdb
> 473.320u 48.230s 12:40.35 68.5% 0+0k 0+0io 273735pf+0w
> total 1048434
> -rw-r--r-- 2 hyc users 98 Jan 26 03:31 DB_CONFIG
> -rw------- 1 hyc users 8192 Jan 26 04:40 __db.001
> -rw------- 1 hyc users 503324672 Jan 26 04:40 __db.002
> -rw------- 1 hyc users 2162688 Jan 26 04:40 __db.003
> -rw------- 1 hyc users 368640 Jan 26 04:40 __db.004
> -rw------- 1 hyc users 16384 Jan 26 04:40 __db.005
> -rw------- 1 hyc users 131985408 Jan 26 04:39 dn2id.bdb
> -rw------- 1 hyc users 431702016 Jan 26 04:39 id2entry.bdb
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
http://www.symas.com http://highlandsun.com/hyc
Symas: Premier OpenSource Development and Support