RE: how do you do this. (philosophy and politics)
- From: "Steve Ollig" <sollig@xxxxxxxxxxxxx>
- To: <niall.litchfield@xxxxxxxxx>, "Oracle-L" <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 11 May 2007 17:03:59 -0500
> 1. how do you define business tasks?
I think the definition of the business tasks is something any business manager
should be able to do. It's why she exists after all - to manage the completion
of said tasks. She should also be capable of prioritizing them. I'd let the
business define & prioritize the tasks. Perhaps use a process similar to the
way you define requirements for a project.
> 2. how do you monitor them?
As Cary is so good at pointing out, response time is best the answer. More
specifically you break them down and monitor response time as granularly as you
need to (sufficient drill-down data). The tasks with the longest response
times are the ones with the most opportunity for improvement. It might be nice
to put a dollar value on response time like you suggest. But I think cost or
revenue impact has more to do with the importance of the task than monitoring
the SLA. Cost & Revenue impact would also have a lot to do with the potential
return on investment for improving response time.
> 3. how do you set target response times for them.
I think I would start monitoring them first and go from there. The SLA should
be stated in terms of the task meeting the target response time X% of the time.
Now I'll have to fess up, I don't have any real world examples of well defined
SLAs. Most places I've worked the SLAs are far more loosey goosey than I'd
like (if they exist at all). I also have the advantage of recently attending
one of Cary's seminars. If you're reading Cary - hope I listened well ;)
So when it comes down to relating your tuning efforts to business performance
targets, I think a well structured SLA will get you there. If you don't have a
good SLA and therefore find yourself tuning something because someone
complained it's too slow (as I often do) be sure it's something that will make
a difference when it goes faster. The only measurement I have for those
successes are the frequency of the complaints.
Anyway - hope those ideas help Niall.
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]On
Behalf Of Niall Litchfield
Sent: Friday, May 11, 2007 2:26 PM
To: Oracle-L
Subject: how do you do this. (philosophy and politics)
In response to a recent question about monitoring Cary replied
Simple answer to this one: Anything that causes increase in business task
response times.
So, if you're monitoring end-user response times with sufficient drill-down
data (such as you can get from your Oracle extended SQL trace data), you're
monitoring what you need to monitor.
I'm guessing there won't be many objectors to that here, and you certainly
won't find me among them if there are. So my question is really the political
1. how do you define business tasks?
2. how do you monitor them?
3. how do you set target response times for them.
I can see in the case of a sales oriented organisation (Amazon, Oracle, ... )
how you *might* relate tasks to revenue - though I reckon this is harder than
might appear. In my case I work for a large public sector organisation with a
number of different internal facing and external facing systems that rely on
Oracle (and that other software provider beginning with M). My last position
was similar. In neither case was I able to get any business person to define
either business transactions or acceptable response time. Costs of downtime
*do* tend to get defined by the way - in cash terms and risk to the lives of
the public for people who really like this stuff.
Now part of this is public sector risk avoidance (it maybe large corporation
risk avoidance to be fair), but I think most of it is down to needing a good
way to describe the impact of response time on business performance, I've yet
to see one. I suspect the same applies to throughput/scalability impact on
business performance but I've not seen any research on that at all.
So my question is, how do other people relate their specific tuning efforts to
business performance targets, and how do they engage non-technical staff in the
process?
--
Niall Litchfield
Oracle DBA
http://www.orawin.info
p.s.
throughout this piece i've resisted the temptation to use the phrase "the
business" since I believe it to be a big part of the problem. maybe I'm wrong.
______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email
______________________________________________________________________
Other related posts:
