Performance profiling through the independent trace facility

Edit online

The performance profile information in trace is intended to help users diagnose performance problems. By using the independent trace facility, performance profiling is accomplished with minimum impact on server performance.

The independent trace facility profiles operation performance that consists of time stamps at key points that are traversed during an operation execution for a running server instance. The following time stamps are profiled during different stages:

RDBM search processing
RDBM bind processing
RDBM compare processing
RDBM write processing

Note: Time stamp collection points for individual operations are provided only for the RDBM backend.

The instance configuration option ibm-slapdStartupTraceEnabled governs the tracing of performance records at server startup. With dynamic tracing (ldaptrace client utility), the independent trace utility can be made to start or stop collecting performance records after server startup. To activate tracing of performance records dynamically, run the following steps:

Activate tracing for performance records. Run the ldaptrace command of the following format:

ldaptrace -h <hostname> -p <port number> -D <adminDN> -w <adminpwd> -l on \
 -t start -- -perf

Dump the trace to a binary trace file. Run the following command:
```
ldtrc dmp trace.bin 
```
Format the trace. Run the following command:
```
ldtrc fmt trace.bin trace.txt 
```

After formatting the trace you can analyze the trace and diagnose performance problems. To turn off tracing, issue the following command:

ldtrc off

Formatted performance trace is shown in the following example:

prf_entry LD PERF FrontEnd::operation_in_workQ (3.100.98.1.1.0)
pid 10255; tid 1167334320; sec 1159183071; nsec 84815000
En-queue bind op; Worker thread ID 1133718448; 
Work Q size now = 1; client conn (9.124.231.39: conn ID 1)