Hi there everybody,
ill put my post in queue here. Hope you dont mind!
I used the following metric areas for my project's performance baseline:
1) Number of transactions/sec(batch request/sec counter within SQL 2008).
2)response times (i have found some counters for queries in MDW). Can you please add something?
3)Throughouts(normal,during peaks) - batch request/sec counter from MDW found in server activity history report.
4)Downtime - i came up only with SQL counter for errors. Is there something else i could employ?
5) Disk utilization - for this i use PerfMon counters. Lots of them...or MDW Disk usage history.
6) the same as above for CPU usage..
7) Database server memory - counters from PerfMon.
8) Query performance statistics - MDW standard report of Query Statistics reporting.
9) Disk latency - counters from PerfMon.
10) Indexing - PerfMon counters for usage of indexes. Plus i think i should include something for index maintenance..when the db grows.
I would be greatful if you could provide some comments on the metric areas given above and add up things that i didnt take in consideration.
Its actually a complete system involving several clustered db servers (SQL 2008), mainframe systems, 2 windows app servers. The communication within the system is based on MQ. Hope that answers your question.
My mentor has actually suggested to use the Appdynamics that allows the end-to-end transaction performance measuring..we just need to find a way for tracking the MQ msgs (maybe extracting the guid of transaction). And i'm currently studying it..but still i need some feedback on the metrics that i have considered in order not to miss out anything.
I would be greatful if you could provide some feedback.
p.s. i am also thinking of employing the MDW (only some counters for db) and maybe PerfMon..
Classical system analysis focuses on resource constaints in four core areas
You may want to examine how your monitors expose resource constraints in these areas on the platform in questionm in your environment.
A typical pairing will include generic OS/hardware metrics for the above items as well as platform/service specific items to understand better how your particular service, such as a database, is managing each of the metrics. The more efficient the use of a finite resource the more scalable the solution.....and you might also add, the cheaper the run in both physical environments (lower hardware needs) and cloud/utility environments where the solution would be charged based upon use of resources.