Search This Blog

Thursday, February 7, 2008

IBM: the Internet really should be a mainframe (NOT)

Nicholas Carr notes that IBM Research has published a paper hypothesizing we should collapse the entire Internet ecosystem into one big computer. Why? Because such a computer could be dramatically cheaper to create and operate than today's commodity clusters.

At present, almost all of the companies operating at web-scale are using clusters of commodity computers, an approach that we postulate is akin to building a power plant from a collection of portable generators. That is, commodity computers were never designed to be efficient at scale, so while each server seems like a low-price part in isolation, the cluster in aggregate is expensive to purchase, power and cool in addition to being failure-prone. Despite the inexpensive network interface cards in commodity computers, the cost to network them does not scale linearly with the number of computers. The switching infrastructure required to support large clusters of computers is not a commodity component, and the cost of high-end switches does not scale linearly with the number of ports. Because of the power and cooling properties of commodity computers many datacenter operators must leave significant floor space unused to fit within the datacenter power budget, which then requires the significant investment of building additional datacenters.
It's a terrifically thought-provoking vision. But its a vision we've heard from IBM in other contexts, when it pushed running Linux on Z-series mainframes as being more economical than buying its own Intel X-series servers. And last time I checked, IBM's X series servers were still selling just fine. Ultimately, customers, not research papers, will really determine the types of servers we use on tomorrow's Internet. But the factors that this paper leaves out are the very real and human issues of control, governance, and freedom. Yes, we need big iron clusters and servers to deliver services to consumers Anywhere. But we'll only see innovative Anywhere services emerge if the companies owning the the big iron doesn't have veto power over entrepreneurs wanting to build those services. And regardless of the economic benefits of a mainframe Internet, tomorrows Anywhere networks will need innovative services more than lower bandwidth and CPU costs.

1 comment:

Anders said...

Totally agreed on the governance point. I would add that resilience is also a major factor. Millions of commodity machines might be inefficient but in aggregate they supply a far more resilient service. The Internet has never been completely down since its inception because it would be nearly impossible to stop all those irregular machines with differing operating systems with their individual strengths and weaknesses. No one virus, for example, is going to take them all down.

Imagine if the undersea cable cut in the Middle East had meant a 100% outage of the one machine...