Egenera Blog

The “Google model” remains a constant touchpoint for most folks in the technology industry. How they do it, why it works, etc. My take has always been, “the Google model of assembling large numbers of commodity servers in a huge grid only works for certain types of applications.” Historically, grids have been really good in the niche - high performance computing types of applications for example. But to be fair, grids have also emerged within the data center to support the service provider market with companies like Google, Yahoo, Microsoft, Amazon, Linden Lab (creator of virtual world Second Life). Regardless of the industry, the only applications that can efficiently utilize a “grid model” are those that handle dispatch problems with no persistent data.

Rackable, which provides many of the servers for this market segment, has done very well for itself, targeting the high-volume data center and HPC markets. The only problem is that its revenues rely on a small set of large customers, its differentiated packaging may not be sustainable, and the market is highly competitive (HP, IBM, Sun and other HPC vendors…Cray, Linux Networx, Liquid Computing). To be clear, these are issues many of us in the data center market share. But no matter how many challenges we have in common, these industry forces are now applying pressure from every side.

Low prices for servers are good for customers (especially those using them within a grid), but the cost for hardware only represents 7% of the total IT budget for large enterprises. Grids would create an unbelievable amount of management complexity if applied to all applications – companies already spend 2/3 of the IT budget on operations and management!

On the other side of the spectrum, business critical and mission critical applications – ERP, CRM, messaging, trading, billing, ordering, and business intelligence – require different models. It is possible to leverage commodity components to deliver computing resources to business and mission critical applications, but a new approach is needed.

And that is where we’re focused.