About frogs, storks and Linux - or: The nature of pseudo correlations
This is really interesting. I´ve talked about this effect at my presentation on Friday. There is an effect in statistics called pseudo correlation. You have an pseudo correllation, when you can proof a correlation between two attributes , albeit the real relation lies somewhere else.
A good example: In years with many storks, more babies are born. And well, we were able to proof this relation in the past. But i hope all my reader are aware of the fact, that babies are not delivered by storks. The real corelation is somewhere else. Animals tend to reproduce them more effectively in years with good conditions. And albeit we try to hide it, humans are nothing than mamals. Okay, in former years wet years were good years for growing food. But wet years are good for frogs as well. And storks like frogs … for lunch. More frogs, more storks. More water, more frogs. More water, more food. More food more babies. Albeit the amount of storks and the amount of babies seems to have acorrelation, it´s an indirect one. Okay, getting from mamals, storks, babies and frogs back to the topic of this blog entry. I found an article in SearchEnterpriseLinux.com: “Move from Solaris to RHEL boosts performance for the Chicago Mercantile Exchange. In the first moment i thought: What the fsck …? But then i´ve read the article and it´s pseudo coreleation time again. The CME had Sun Server in the past. These Sun systems operated with Solaris. They purchased new servers. They use RedHat on them. Okay, that´s bad for us, but hey … that´s life. The strange part are the conclusions in this articles. What would you think, when you read the headline. “Solaris is so slow … and Linux is much better”. I want Linux. I want this boost, too. When you read the article, you come to a different conclusion: Their baseline were Sun servers from 2000-2003. That´s 8 years ago. In IT this 32 hype cycles (a hype a quarter … good rule of thumb) in the market and and least 3 real technology cycles (UltraSPARC III, UltraSPARC IV and IV+, Sparc64 VI and CMT) at Sun. They´ve redesigned their business logic. And obviously their application got faster by the help of Mr. Moore. And now i come to the big pseudo correlation:
The migration to Linux has paid off in improved performance, lower cost and greater stability, Kutty said.
Mr. Kutty … i don´t want to insult you, but this is nonsense. Your performace boost didn´t came from Linux, it came from using faster servers and redesigning your business logic. At least this is the real conclusion out of this article. I´m pretty sure, that with Solaris under the hood you would have yielded the same effect with new hardware and the rewritten application. But wait … when you really read the quote, he is pretty aware of the fact, that it wasn´t Linux that gave him the boost and that it was an effect of the migration instead. But SearchEnterprise linux got this totally wrong. So, what´s the sense behind this story? There is none … i assume, at least in regard of the headline.Well … pseudo correlations are still a standard tactic of marketing and to answer the “Let us known what you think about the story”. Well, Mrs. Derringer … i have to admit .. nothing good.