Sybase 12.5.1 User Manual page 185

Using Process ID
Using Procedure Elapsed Time and Procedure CPU Time
User's Guide
APPENDIX C
When a server process terminates, Adaptive Server can reuse its Process ID for
a new process. Therefore, the Process ID data item is not guaranteed to
uniquely identify a process. The Kernel Process ID data item, however,
uniquely identifies a process.
To create a non-raw playback view that shows per-process data, the input view
and the playback view must include both the Process ID and Kernel Process ID
data items. If playback is raw, using only Process ID is allowed.
Views that include Process ID return rows as follows:
• Recording session views and raw playback views.
These views return a row only for process IDs that exist at the end of a
sample interval. If a server process terminates in the middle of a sample
interval, a row is not returned for its process ID.
• Playback views when the session's summarization_level parameter is
, , or user-defined intervals.
actual entire
These views return rows for all server processes included in any of the
input views. However, since the Process IDs are not guaranteed to be
unique, the Kernel Process ID data item must also be included to ensure
uniqueness of the key. Otherwise, the view could erroneously summarize
two different processes.
This section discusses how to use the data items that can have the "Avg for
Sample" and "Avg for Session" statistic types. The data items are:
• Procedure Elapsed Time
• Procedure CPU Time
If the playback view uses these data item names with the "Value for Sample"
or "Value for Session" statistic types, the input view must have the same
statistic type.
If the playback view uses these data item names with the "Avg for Sample" or
"Avg for Session" statistic types:
• The returned value is calculated if the input view includes the additional
data item Procedure Execution Count.
• The returned value is estimated if all of the following are true:
Specifications for Defining Playback Views
171