Michael T. Nygard Quotes

Things happen in production - bad things that you can't always predict. One

True utility computing centers are on the horizon, but right now, the only real ones are a pale approximation of this vision. In the world that the other 99.9% of us inhabit, production systems are deployed to some relatively fixed set of resources. Applications

The silicon microchips themselves might be cheap (relative to times past, anyway), but CPU cycles are not cheap. Every CPU cycle consumes clock time. Clock time is latency. A wasteful application makes its users wait longer than they need to, and if there's anything users hate, it's waiting. For web systems, latency in the application has a dual effect. The added processing directly increases the burden on the application servers themselves. Suppose that an application takes just 250 milliseconds of extra processing per transaction. If the system processes a million transactions a day, that extra 250 milliseconds per transaction makes for an extra 69.4 hours of compute time every day. Assuming an 80% load factor on each server, you'll need four additional servers to handle this load.

Design with skepticism, and you will achieve resilience. Ask, "What can system X do to hurt me?" and then design a way to dodge, duck, dip, dive, and dodge whatever wrench your supposed ally throws.

One of my favorite retailers has a release process that rivals a NASA launch sequence.

Software design as taught today is terribly incomplete. It talks only about what systems should do. It doesn't address the converse - things systems should not do. They should not crash, hang, lose data, violate privacy, lose money, destroy your company, or kill your customers.

See whether the DBA laughs at the queries If it doesn't pass the laugh test, it shouldn't go into production. Period.

Keep reports out of production Reports can, and should, be served elsewhere. Don't jeopardize

of The TCP/IP Guide [Koz05] or TCP/IP Illustrated [Ste93] open beside you for this type of activity!

Integration points are the number-one killer of systems. Every single one of those feeds presents a stability risk.

Most testers I've known are perverse enough that if you tell them the "happy path" through the application, that's the last thing they'll do. It should be the same with load testing.

First, nothing is as permanent as a temporary fix. Most of these remained in place for the next year or two.