During our performance evaluation, we run one load simulation with SQL trace on. And while analyzing the trace file I want to calculate the CPU usage. How to do it? (Please refer the screenshot attached)
So for the physical read, we observed a CPU spike but don't know how to interpret it.
Tha value showing for 53,682,888 reads, the CPU utilized is 1,254,911. Is it in a microsecond or in cycles?
These values are represented in terms of time. This value could either be in millisecond or microsecond. For CPU - it would be number of milliseconds used by this particular process.
This has been explained by Mr. Brad McGehee in a very detailed way with examples.
CPU
This data column shows the amount of CPU time used by an event (in milliseconds). Obviously, the smaller this number, the fewer CPU resources were used for the query. Note that CPU data column is only captured for the RPC:Completed, SP:StmtCompleted, and the SQL:BatchCompleted events.
Reads
This data column shows the number of logical page reads that occurred during an event. Again, the smaller this number, the fewer disk I/O resources were used for the event. Note that Reads are only captured for the RPC:Completed, SP:StmtCompleted and the SQL:BatchCompleted events.
Writes
This data column shows the number of physical writes that occurred during an event and provides an indication of the I/O resources that were used for an event. Again, Writes is only captured for the RPC:Completed, SP:StmtCompleted and the SQL:BatchCompleted events.
Please find the below notes from Microsoft on this:
Note
The server reports the duration of an event in microseconds (10^-6 seconds) and the amount of CPU time used by the event in milliseconds (10^-3 seconds). The SQL Server Profiler graphical user interface displays the Duration column in milliseconds by default, but when a trace is saved to either a file or a database table, the Duration column value is written in microseconds.
Same can be summarized here:
I hope above helps.
