The Perils of Platform Selection
By Ray Valdés
The subject of Web site platforms and architectures is one that is of compelling importance to some site managers and excruciatingly boring to others. This is partly because when a platform truly works well, it becomes almost invisible. It seems that many people only notice platforms when something goes wrong.
The problem, of course, is that once things go wrong, the effects can be dramatic, and the tarnish on a site's image can linger for a long time. Witness the well-publicized eBay outages that shook loose almost $10 billion out of the company's stock market valuation. Since the low point last summer, the company's stock price has climbed slowly but has not yet reached preoutage heights.
When a platform malfunctions, there isn't always an easy fix. At one point in time, it appeared that the Sun E10000 machines were somehow involved in the system outages at eBay. The only alternative for eBay management was not to back out of that platform, but rather to buy a bunch more of these million-dollar machines, for added redundancy and hot backup. (For the record, Sun and eBay have backed off from earlier statements implicating the E10000 systems.)
A platform selection decision is often a till-death-do-us-part commitment for a business that runs deeper than most marriages. Users and developers usually spend more time with their platforms than with their loved ones. When things go awry with a platform, often the only available response to tribulation is not to abandon the relationship, but rather to get counseling (such as consulting high-priced experts) and attempt to make things work with the chosen one.
