Wait On Seamphores Are High

[Jack@dba]

Hi Team,

Today we found in performance reports that wait on semaphore is getting high in number.
For last 3 months values as follows 810.00 , 840.. and 900.00.

We will take this values on monthly basis by taking overall average value of whole month.

FYI..........

-n 140
-Mn 51
-spin 200000
-semsets 6

IBM AIX 7.1 and openedge 11.6 enterprise edition.

Why wait on seamphores are increasing? How to tune it?

Thanks
Jack

 

[TomBascom]

My first guess is that your workload has increased. But I'm just guessing because you have shared hardly any data to work with.

What possessed you to set -spin to 200,000?

Your wait on semaphore numbers lack context -- is that total waits? waits per second? waits per transaction? something else?

Why do you say that that number is "high"? "High" compared to what?

Averaging over an entire month seems unlikely to be useful. Most people have significant day/night and workday/weekend components of their workload that would make that approach useless. What is you rationale for taking this approach?

Other than semaphore waits being values that offend you -- is there an actual problem with tangible symptoms? Are there measurements of this problem that might be used to describe it?

 

[Jack@dba]

Your wait on semaphore numbers lack context -- is that total waits? waits per second? waits per transaction? something else

Ans) we have script which run every on monthly Ang get the details of all promon Information on monthly basis.the values which I provided are average of total promon files from. 1st to 30th.

Why do you say that that number is "high"? "High" compared to what?
The values for semaphore waits are gradually getting increasing as per our reports.

Averaging over an entire month seems unlikely to be useful. Most people have significant day/night and workday/weekend components of their workload that would make that approach useless. What is you rationale for taking this approach?

We will prepare performance report of every month using promon details.

Other than semaphore waits being values that offend you -- is there an actual problem with tangible symptoms? Are there measurements of this problem that might be used to describe it?

As of now there is no problem from database or application end but I want to know how to tune this situation?