I was about to enter a meeting yesterday, when a peek into the room showed me everyone dressed “business casual.” Seeing as I was wearing a suit and didn't want to be overdressed, I flipped off my shoes and headed on in.
It wasn't long before someone commented on my suit. Thankful for ability to think on my feet, I pointed to my socks and said, “I realized the suit was a bit formal so I ix-nayed the footwear... on average, I'm business casual.”
When the clamor in the room subsided, the general consensus was that I was insane.
And, actually, that was my point. See, I wore the suit on purpose, though I admit, I didn't think of the shoe-schtick until I was outside the room. My point was...
It's insane to make important decisions based upon averages.
SOA What? You have a service. It's used by several applications. You want to know how that service is performing. What do you measure?
Well, you might measure service availability or response time. You might measure how much CPU it's using, or how many messages per second it can process. But, you never consider the application's many contexts. Why's that?
Well, because it's hard to do that.
Does a manufacturing system sharing a service with an HR time-clock system also share performance requirements? Does a market research application have the same performance characteristics as a trading application? If average SLA compliance is OK, why isn't average dress-code compliance OK? Our dress code is, after all, just another “business policy.”
Service owners often are unaware of the contexts in which their services are used. In fact, that's a key SOA driver! Sharing services without worrying the implementation details. The problem, in these shared services examples, is that the “on average” is implied. Though, clearly, averages are unsuitable measuring compliance.
Please, if you don't want your sanity taken in question like mine, examine your own SOA compliance with business policies for exposure to averages... before it's too late.
*I realize that, technically, on average half of your services are performing worse than the median. But, that didn't sound as catchy.