Just a quick question, as I was pondering today - what is good performance? I was thinking there may only be designations such as "acceptable" performance, or poor/bad/unacceptable performance. Is acceptable what you would ordinarily define as good, or is there a level above acceptable? And, if there is a level above acceptable, do you want to obtain it (or would that be premature optimization)?
Thoughts?
Hey! Why don't you make your life easier and subscribe to the full post
or short blurb RSS feed? I'm so confident you'll love my smelly pasta plate
wisdom that I'm offering a no-strings-attached, lifetime money back guarantee!
Leave a comment
I'm guessing acceptable (when quantified) would be done so in terms of a range. If goal is to meet contractual obligations as quick and cheap as possible "good" performance would be as close as possible to the bottom of that range, if you wanted a REALLY happy customer, good performance could be defined as anything on or above the top bound of the range and to aim for that.
It's also all about stats as you're gonna see a gaussian distribution, so a given performance number would have to be beaten a percentage of the time excluding downtime events.
Where you want to be depends on what your objectives are. If it is a business, you'd be looking for ROI based on good will, competitive differentiation, safety margin (so you don't get phone calls if only 1 of your servers in your web farm crashes - one extra server? $1200. Being able to sleep through the night and replace the HD on Monday? Priceless!) and the like.
Posted by
Peter Bell
on Mar 30, 2007 at 12:34 PM UTC - 6 hrs
"I'm guessing acceptable (when quantified) would be done so in terms of a range."
Well, that's what I was getting at. Would you ever say anything above that range is unacceptable? I guess that might happen- for old programs that based timing on the number of cycles (I've played some games that just go way too fast to be of any use on newer processors).
But in most cases, I thought anything above a certain threshold, since I wouldn't expect better than acceptable to ever be unacceptable.
In any case, thanks for the detailed treatment of some of the options available - I hadn't quite thought of it in those terms before.
Posted by
Sam
on Mar 30, 2007 at 01:12 PM UTC - 6 hrs
I'd say it would be unacceptable if it cost you more money (resources, time, opportunity cost, etc) and didn't provide a better return than the best business case for otherwise using those resources.
Can you TELL I'm in "Pointy Haired" mode today?!
Posted by
Peter Bell
on Mar 30, 2007 at 01:14 PM UTC - 6 hrs
Well that is certainly true. But would that be a case of the application performance being unnacceptable, or the decision to squeeze blood from a stone?
And clearly, I'm trying to split those pointy hairs. =)
Posted by
Sam
on Mar 30, 2007 at 01:20 PM UTC - 6 hrs
