Data Integration Blog -

Calxeda has gotten a lot of coverage over the last few days about their strategic partnerships, and being within a year of generally available ARM-based servers (Austin American Statesman, Channel Register, GigaOm, MSP News). One of the partners they announced is the company that I work for, Pervasive Software, which makes Pervasive DataRush™, a revolutionary platform for developing applications that optimally parallelize processing of huge data volumes in, as my CTO Mike Hoskins says, “Small time, small dollars, and small energy.” I’m not on that product team, but my inner geek squeals with glee every time I hear about the things they’re doing over there. The DataRush folks have been working with Intel,  AMD, and now Calxeda, developing amazing multicore data crunching performance at a fraction of previous energy requirements. A lot of emphasis gets put on the performance speed advantages of DataRush, but it’s probably the energy savings aspect that Calxeda is focused on. And to the world at large, I believe that’s far more important.

A lot of folks know I’m a data management/data integration/data quality-focused kind of person. What most folks in my professional life don’t know is that I’m also a bit of a “tree-hugging granola chick,” to quote my husband. I’ve known for a while that data centers in most enterprises are hopelessly inefficient energy hogs, but up until now, there haven’t been any good alternatives that can do the heavy lifting required — just like with cars.

I drive a car that runs on plain old-fashioned fossil fuels, despite a strong desire to drive something more environmentally friendly. Up until recently, there weren’t any practical alternatives for a mom in Texas where everything is far apart, with kids and camping gear, groceries and bicycles to transport. That’s changing, finally. Better hybrids are just now becoming more generally available. Five years ago, when I went to buy a car, there was a nine month waiting list for a Prius, and that was the only viable green option for someone who needed the power and space to haul 5 people and their stuff up a hill. Now, there are a variety of hybrid, alternative fuel, and electric options, even hybrid SUVs with 4-wheel drive for green folks in Colorado where everything is up or down. Car manufacturers are finally starting to catch up to the demand for greener car options for everyone.

Data centers in the United States alone use in the neighborhood of 50 billion MegaWatt/Hours every year, a carbon footprint equivalent to 50 million cars (from GigaOm interview with Calxeda VP). Just as with cars, the demand for better options has been there for some time. In some cases, the main limit on data center expansion in enterprises is not space or money, but power available in that area. The data that companies need to process keeps getting larger by leaps and bounds, 60% a year according to IDC. But corporations can’t afford to grow their data processing centers by 60% a year, every year, and the earth certainly can’t afford for energy consumption and the carbon footprint of data centers to grow by 60% a year.

It isn’t an option to simply skip processing all that data. The flow of data is the life and breath of a healthy modern company. Recently, cloud computing has become a nearly ubiquitous corporate movement. There are a lot of drivers for that, but one of those drivers is clearly the ability to process more data more efficiently, using the giant, elastic, optimized stacks of the cloud data centers. Many companies feel like they can solve their growing data processing and energy consumption problems by moving them to the cloud.

But the cloud isn’t a long-term energy solution, just a stop-gap. Cloud computing isn’t some magical data-in-the-sky thing.  It simply means that data is being processed in someone else’s data center, or if it’s a private cloud, in a re-vamped version of your own data center. The cloud platforms are still running on the same, inefficient, energy hog X86 hardware, although more efficiently utilized.  There may be hundreds of blade servers or dozens of super-computers with the load distributed across them, but they’re still burning power at a high, and constantly growing, rate.

On the other end of the spectrum of efficiency are cell phones which are also ubiquitous these days, and becoming more and more sophisticated. I read recently that a modern iPhone has the equivalent compute power of an ‘80’s era Cray super-computer. The salient point, though, is that it provides that level of data crunching on one tiny battery, for a full day or more. That’s stunning energy efficiency. And it’s largely due to the ARM processor inside.

Certainly an ARM processor from a cell phone doesn’t have the same level of compute power as a standard X86 server, but Calxeda is optimizing processing speed with technologies like DataRush, and partnering with folks like Eucalyptus to provide cloud data center ready hardware. The ARM processor is so efficient energy-wise, that when everything is put together to reach equivalent performance levels to current servers, it will still use 1/10th of the power.

That’s what Calxeda is working on, new chips and servers based on ARM that combine the compute power of super-computers with the energy efficiency of a cell phone.

They’re expecting 50-80% reduction in data center energy usage for the same level of performance, and they’re talking about having servers with that capability generally available in a year.

It can’t come soon enough in my opinion. I know that this will save companies that use it money in the long run, and undoubtedly make Calxeda and their partners money, too. But the real value here isn’t monetary. It’s chopping down that carbon footprint, and keeping it at a manageable level for the foreseeable future for all us tree-huggers out there, and anybody who likes breathing clean air. The world needs more tech like this.