DPA Overhead Test

Database Performance Analyzer was a hard sell to my DBA team. Even once we bought the product the monitoring team was not allowed to connect it to our production Sybase ASE databases for fear that it would cause too much "load" and negatively impact performance. The monitoring team conducted a thorough test and I wanted to share the results.

The test was conducted using Sybase ASE 15.7 running on AIX 6.1. DPA version is 12.0.3074 running on Windows 2016 and SQL 2016 repository.

We would prep the database for our largest batch job. snap it, run the batch job, revert back to the snap and run the batch job again with DPA connected.

These are the results. I realize the context of the test won't apply to most but the delta from DPA enabled/disabled is what really matters which ended up being approx 2.2%

Metric	DPA Enabled	DPA Disabled
Average Claims Per Hour Per Engine	952	973
Claims Per Hour	60985	62287
Claims Per Minute	1016	1038
Step 4000 Duration	125 mins	122 mins
Total Claims	127274	127274

Find more posts tagged with

aix

Performance

test

overhead

load

dpa

sybase

ase

Accepted answers

All comments

darichar

Brian,

Really interesting data, thanks for sharing. Mind if I ask a few clarifying questions:

The Total Claims number is the same for enabled/disabled even though the claim metrics in the first 3 rows are lower. Is that a typo?
Is "Step 4000 Duration" part of an internal custom process?
Are these numbers an average of many iterations of this process? If so, how many did you perform?

Thanks,

Dean

muckman

Dean,

Total claims are the same to show the volume of the two runs were identical. The batch job processes the data set until complete. The speed at which it processes is something that is highly tuned and monitored. "Step 4000" is the biggest job that takes the longest of the entire workflow and the one selected for the test, its also overall duration time for the run. We only ran this test twice. I wish we were able to run the test again to back up the data but the resources to pull this together was somewhat involved. The batch admins tell me with the same data set they can re run the same job with the same parameters and get the same result consistently over and over. The only change was DPA.

Thanks for the interest.