I’ve seen a number of reports about Internet collapse and a near term data centre collapse. There’s not been this much concern since the mid 90s. Don’t get me wrong some good will come of it, since the only way to prevent problems like this is to share the symptoms and remedies. It does seem to be interesting timing though, since it always seems to happen about this time of year.

If the Internet backbone were to become overloaded in some catastrophic way where it could not be depended upon, the plans of many highly values companies that deliver all their services over the Internet would be in jeopardy. (You know who you are!) On the other hand they have enough market capitalization that they have the deep pockets to apply to the problem. Right now most companies that provide these services only provide service level agreements within their data center and not from an end user’s perspective. All this talk may change that.

Besides the issue of the overwhelming the Internet through the use of data and power there is also concern about a well targeted terror attack. There have been studies about losing a major Internet hub for years. All of these issues though are well within the domain of an organization’s chief risk officer and business continuity planners. Many scenarios can be planned for and tested against. After all, we’ve seen widespread power outages and zero day attacks before.

Now we have Sun warning of Data center collapse. Yes, there are more and larger data centers. Statistically it is a good bet that one will collapse soon. Someone always wins the lottery. Sun VP Subodh Bapat warning: "You'll see a massive failure in a year. We are going to see a data center failure of that scale."

According to Sun CTO Greg Papadopoulos, 2008 will mark the first time companies that virtualize large machines into a parallel computing resource will be a major consumer of compute power. This shift is what Sun has been calling Red Shift. "We're talking about 50,000 sq. ft, 5,000 kW centers now and 500,000 sq. ft, 50,000kW centers over the next ten years."
 
These are some good sized data centers, but nothing that can’t be tested on a regular basis (and should be tested). When I looked at the comments to some of the articles and blog posts about data center collapse, there was a great deal of concern about the size of the power consumption number – for good cause – but I didn’t see anyone talking about how much more efficient these consolidated data centers are when compared to a highly distributed approach that is a legacy of the .COM era. Yes the power requirements are high, but it should be less than having the various applications and related systems spread out everywhere and underutilized.

We should take all these stories into account and dust off that disaster recovery plan and test it (both paper tests – to make sure we’ve got things covered as well as actual tests to make sure they can be executed). 

Charlie Bess
EDS' Next Big Thing Blog