How to Model Transaction Throughput?
We have a large scale-out application that less focused on the elapsed time for a specific business transaction and more focused on overall transaction throughput (i.e. 250 TPS, 300 TPS, etc.). They would like to graph that their TPS load is, and then put SLAs on it like 'severe if less than 300 TPS during a specific time window', 'warn if less than 350 during a specific time window'. This is for a set of overnight cycle jobs that usually take 3-4 hours to run, and a throughut drop would mean that this component won't meet its completion SLA.
Any guidance on how can we implement that in dynaTrace?
Thanks, Michael.
Hi,
In order to chart TPS you can simply use the count aggregation for any Business Transaction in the chart. The count would be on a per resolution basis. Smallest resolution is 10 seconds, making the 300 TPS = 3000. for longer frames the resolution would increase at least to a minute and the count would increase as well. 18000 transactions a minute equal 300 TPS.
For incidents, you should be able to define that the count for a BT is bigger than x for a specific time frame.
e.g. if the count for BT y is less than 18000 for one minute raise a severe incident
This would check every 10 seconds if the number of transactions in the last minute was less than 18000 (300 per second) and raise a severe incident if that is the case. It would close the incident once that condition is no longer met, meaning if the the incident sees more than 18000 transactions in a given minute.
You can of course do this for longer timeframes as well.
Let me know if this helps
