Here are four beliefs that have served me well in every type of performance work I’ve ever done. They are not technical, but they are fundamental to your success. Get this stuff wrong and you will suffer.
All of your efforts will be meaningless and wasted if the people you are working with don’t trust you, the quality of your work, and the results you’ve generated. They bet their job, and to some extent the future of the company, on your work. To succeed you should be honest and trustworthy as a person, and you need to check your work carefully every time as even small errors erodes confidence in your work. Never allow yourself to be forced into lying about the numbers as you present your way up the power chain to more and more senior management.
Never Serve Undercooked Ideas
Once you begin to work on a performance problem, people will almost immediately ask you your opinion on the cause of the problem and lots of questions that boil down to: “Will we be OK?” Nobody wants to wait. I can’t blame them.
Reporting early results, before you’ve double-checked your work, is the road to disaster. You wouldn’t give a hungry guest raw chicken; don’t give hungry coworkers your raw, unchecked, conclusions.
Almost every time I have let someone wheedle an unchecked conclusion out of me it has turned out poorly. Doing this will erode organizational trust in your work as usually your conclusions change as you get more data. All they will remember is that at first you got it wrong. When relentlessly pressed, give them a juicy bit of observable data (example: We are seeing 4x the normal load.), but keep your unchecked and undercooked conclusions to yourself.
Give Outside Experts Some Privacy
If you have outside experts come in to work a problem, it can be tempting to hang around them all the time to learn new things. However, it is in your own best interest to periodically give them a chance to confer privately and give them a private space to talk to the their home office experts.
Why? Out of your earshot they can confer frankly about your problem and hopefully come up with better solutions by:
- Talking freely to each other about what they do and don’t know about your situation.
- Brainstorming without worrying they look like idiots in front of you.
- Speaking candidly to their home office technical wizards about what they’ve observed, what they don’t know, and their current guess as to the root of your problem.
Put yourself in their shoes. If you had to fix a big problem at a key customer, wouldn’t you want some privacy as you sought advice, searched for answers, and evaluated possible solutions?
Always Guess Before Reading a Meter
As you work with any meter I urge you to always take your best guess at what the meter will report before you look at the actual results. Every time you guess the value you sharpen your ability to instantly know something is wrong, or all is well. Unless you are first learning to use a new meter, you have no business evaluating the output of the meter if you don’t have some idea of what would be a reasonable value for it to return in the current situation.
If you just blindly accept what the meters tell you, then all manner of errors can creep into your world, and worse, you will have no clue when they do. If you expect to see 500 reads and you see (20 or 5000) then that is a mystery that needs to be solved. Your test could have crashed early; someone else could be on the machine without your knowledge; or, as I’ve seen in the past, something as simple as an automated backup kicks in and messes up your work. Your data has to make sense to you and be internally consistent.
If you have no expectations about the results of your metering and how these meters relate to each other and the transaction load, then someday you’ll take bad raw data, then distill it into a beautiful (but meaningless) report, and give a bogus presentation that might just end up hurting the company and ending your career.
There is more to learn about performance work in this blog and in my book called The Every Computer Performance Book, which you can find on Amazon and iTunes.