HP-UX Directory Server 8.1 Performance Tuning and Sizing Guidelines

Step 4: If the currententrycachecount is less than total number of entries in your database, and

the maxentrycachesize has not reached the maximum value that is available from your system,

increase nsslapd-cachememsize, and repeat from Step2 until all the entries are cached or

nsslapd-cachememsize has reached the maximum value from your system. If

nsslapd-cachememsize has not reached the maximum system value when all the entries are

cached, then set nsslapd-cachememsize to be what you got for currententrycachesize

from step 3. Allow some safety margin if possible.

To estimate nsslapd-cachememsize for your database, you can use the following equation

derived from performance testing results.

Note: Even though the equation has been very reliable in performance tests when tuning nsslapd-

cachememsize for different sizes of databases, you should always confirm the equation’s

nsslapd-cachememsize value with testing before going into production.

Equation 2: Estimate Size of Entry Cache for Performance Testing:

nsslapd-cachememsize = total_number_of_entries_in_the_database *

average_space_each_entry_needs_in_the_entrycache

average_space_each_entry_needs_in_the_entrycache can be found by:

Step 1: Set nsslapd-cachememsize to an estimated value that is large enough to give a good

sample, and set nsslapd-cachesize to –1.

Step 2: Prime the server by executing

ldapsearch command 1.

Step 3: Find out the value of currententrycachesize and currententrycachecount by

executing

ldapsearch command 3.

Step 4: average_space_each_entry_needs_in_the_entrycache =

currententrycachesize / currententrycachecount

Note: For more information about attributes such as currententrycachesize and

currententrycachecount, see the

HP-UX Directory Server administrator guide.

Figure 3 shows performance results using the Montvale-based test configuration (8 CPUs@1.6GHz)

measured against different percentages of database entries in the entry cache.