Archives of the TeradataForum
Message Posted: Mon, 17 Mar 2003 @ 16:18:40 GMT
| Subj: || || Re: Canary queries and SQL for performance monitoring |
| From: || || Meade, Cornelius |
One way to approach what you seem to be asking is through the use of a canary query. This is a query (or maybe a couple of queries)
which is somewhat representative of your typical business transaction and which you can execute repeatedly throughout the day. By measuring
the response time for these executions at different times of the day and comparing them to norms for your environment (i.e. you also need to
track the overall system workload during the day) which equate to system performance within the bounds you deem acceptable then you can
probably extrapolate that future execution metrics collected for the canary query mean the system performance is within the bounds
previously established. This is basically how I've used "canary queries" in the past to establish a dashboard for system performance.
Ideally, you want your canary query to be submitted and executed in as close to the same manner as all other business transactions are
submitted but often this is where some tradeoffs have to be made. Mind you, this type of thing is not fool proof and can be labor
intensive to maintain its validity as a metric but this can work reasonably well.