1
# $OpenLDAP: pkg/openldap-guide/admin/tuning.sdf,v 1.8.2.1 2006/01/03 22:16:03 kurt Exp $
2
# Copyright 1999-2006 The OpenLDAP Foundation, All Rights Reserved.
3
# COPYING RESTRICTIONS APPLY, see COPYRIGHT.
7
Note: this chapter needs to be updated to discuss BDB tuning.
9
There are several things you can do to tune the performance of
10
slapd for your system. Most of them have to do with the LDBM
11
backend. LDBM uses an index mechanism to store and retrieve
12
information in slapd. Each entry is assigned a unique ID, used to
13
refer to the entry in the indexes. A search for entries with a
14
surname of "Jensen", for example, would look up the index entry
15
"=JENSEN" in the surname index. The data returned is a list of
16
IDs of entries having that value for the surname attribute. We
17
have found several things to be useful in improving the
18
performance of this indexing scheme, especially on modify
23
H2: The allIDs threshold
25
Some index entries become so large as to be useless. For
26
example, if every entry in your database is a person entry, the
27
"=PERSON" index entry in the objectclass index contains every
28
entry. This returns very little useful information, and can cause
29
significant delays, especially on updates. To alleviate this
30
problem, we have introduced the idea of an allIDs index entry.
32
The allIDs entry stands for a real index entry containing the IDs
33
of every entry in the database, but it takes up very little space,
34
never needs updating, and can be manipulated quickly and
35
efficiently. The trade-off is that it does not prune the set of
36
candidate entries at all during a search. This must be done
37
using other, more "high-powered" index entries.
39
You can set the minimum number of IDs that an index entry may
40
contain before it turns into an allIDs block by changing the
41
{{EX: SLAPD_LDBM_MIN_MAXIDS}} variable in the
42
{{EX: include/ldapconfig.h}} file. The actual number is determined at
43
runtime by the LDBM backend, depending on the block size of
44
the underlying device (i.e., the number you provide is rounded up
45
to the nearest multiple of a block size).
51
The LDBM backend can be configured to keep a cache of
52
entries in memory. Since the LDBM database spends much of its
53
time reading entries from the id2entry file into memory, this cache
54
can greatly speed performance. The trade-off is that the cache
55
uses some extra memory. The default cache size is 1000
56
entries. See the discussion of the cachesize option in Section
57
5.2.3 on LDBM configuration.
63
The LDBM backend uses a number of disk-based index files. If
64
the underlying hash or B-tree package supports in-memory
65
caching of these files, performance can be greatly improved,
66
especially on modifies. The size of this in-memory file cache is
67
given by the dbcachesize option, discussed in more detail in
68
section 5.2.3 on LDBM configuration. The default {{EX: dbcachesize}} is
73
H2: Maintain the right indices
75
Finally, one of the best performance tune-ups you can do is to
76
make sure you are maintaining the right indices. Too few indices
77
can lead to poor search performance. Too many indices can
78
lead to poor update performance. For example, the LDBM
79
backend would be perfectly happy to maintain substring and
80
approximate indices for the {{EX: objectclass attribute}}, but this would
81
not be useful and would just slow down update operations. If
82
your database has many entries and is handling queries for
83
substring equality on the surname attribute, you should make
84
sure to maintain a surname substring index so these queries are
87
So, take a look at the index lines in your slapd configuration file to
88
ensure that only those indices that make sense and are needed