Migrating Sun Java Directory Server to HP-UX Directory Server White Paper

calculating the disk space required for HPDS. In the following example, the disk space used by the

SJDS instance directory is calculated as approximately 448 MB:

# du -ks /export/instance1

448104 /export/instance1

Note:

The file system paths for the database can be customized. If the default

locations are not used, then you should query each path in addition to the

instance directory. In the following example, the paths for the database

environment, transaction log, and “dc=example,dc=com” suffix are

discovered:

# dsconf get-server-prop db-env-path

# dsconf get-server-prop db-log-path

# dsconf get-suffix-prop dc=example,dc=com db-path

Disk space usage for the database can vary depending on the database configuration settings. By

default, HPDS uses an 8192 byte page size for the database, as is configured by the nsslapd-db-

page-size parameter found in the dse.ldif file. Depending on the configuration used in SJDS,

disk space requirements may increase when moving to HPDS. If the default page size is used on

SJDS, HP recommends altering the nsslapd-db-page-size parameter to a value of16384 on so

that the resulting HPDS database will yield similarly sized database files. If default values are used

plan for a larger database file nearly 30% over what the SJDS currently consumes.

Memory Usage

An HPDS instance that uses default options only requires about 256 MB of memory. However, extra

memory is required if larger entry caches are configured. To verify the sizes of the entry caches in

SJDS, look for the nsslapd-cachememsize settings. The following example shows that a 1 GB

cache is configured:

# egrep –i cachememsize /export/instance1/config/dse.ldif

nsslapd-cachememsize: 1073741824

Because of heap fragmentation and small block overhead, the memory used by entry caches is

usually larger than the configured value. Plan for the actual memory used by the entry caches to be

around 50% larger than the configured size. Heap fragmentation occurs whenever the cache fills up

and entries are evicted, leaving "holes" in the memory allocated by the directory server. Small block

overhead occurs when a block of memory is allocated and the actual size is rounded up to be

divisible with the smallest block size provided by the memory sub-system.

In addition, more memory is used if multiple memory arenas are configured. Eight arenas are used

by default by the directory server instance, but this number can be tuned using the _M_ARENA_OPTS

environment variable. For more information about tuning _M_ARENA_OPTS, see the malloc(3C)

manpage.

When using multiple arenas, configure a small entry cache, monitor memory usage during normal

operations, and tune the size of the cache to be larger if there is need for increased performance and

sufficient free memory in the system.