I was recently with IBM in Mainz discussing data centre challenges for the 21st century. Interestingly, one of the issues under discussion was about having a green data centre.
Now, environmental friendliness is very much on every politician’s agenda, with everyone trying to outdo the opposing candidate on how green they are – in terms of recycling waste, cutting energy use, and creating fewer carbon emissions by not using cars and planes etc whenever possible. And if people are recycling at home, turning down the thermostat on the heating, and cycling to work, it makes sense for them to look at being green in the work environment too.
Now this is where the problems start. Hands up anyone who can define what is meant by a green data centre. We all know what we think it means, but it is quite hard to come up with a definition that is worth including in a dictionary. And in many ways, it is impossible to have a green data centre because of the amount of energy needed to create the processors and data storage devices in the first place, the amount of energy necessary to run them so that we’re getting decent processing speeds, and the energy required to do something environmentally friendly when the hardware is passed its best and being shipped out.
At the meeting in Mainz, IBM was suggesting ways that the data centre could become greener – by which they meant more energy efficient. They were specifically talking about blade servers rather than mainframes, but I guess most sites have a mixture of technologies and this will apply to them.
IBM made a statement that I found quite startling, but everyone else in the room nodded sagely, so I guess it’s true. I suppose it’s my mainframe background that made the idea seem so strange – you’ll probably be saying, “of course, everyone knows that”. They suggested that the average usage of an x86 server was around the 20% mark and that people were likely to go out and buy another server when they hit 25% usage. This shopping expedition wasn’t necessarily caused by the increased server utilization, it was just the sort of pattern that they had observed. That means these companies would end up with rooms full of servers with around a quarter utilization. The first way to become more green is obviously to get rid of half the servers and double the utilization figures. But how can you do that? Well IBM is very keen on virtualization (and why wouldn’t they be, having been using VM for forty years?). Obviously, virtualization does use slightly more power on a single server than not virtualizing, but significantly less than running two servers.
Their other greening strategy was in the way the blades are cooled. Apparently air conditioning warm air works better than air conditioning cooler air! They told horror stories of servers that were drawing lots of power and were then running hotter, so the fans would spin faster to cool them down, which meant that the fans were drawing more power (and creating more heat, which meant…). Their solution simply involved keeping the hot and cold air separate, which results in the air conditioning working more efficiently and less energy being used.
They did also have ways of water cooling the doors of blade servers to keep down the temperature and some of the components were now more energy efficient, which meant they were greener – although this was a consequence of a desire to make them more efficient rather than anyone specifically following a green agenda.
So, to answer my question in the title of this blog, Big Blue is moving towards greenness. It’s doing it because it makes sense for them to do so. And that’s because energy efficiency means customers can save money – always a strong selling point. And also because customers are asking for greener solutions at affordable prices, which IBM is able to provide. However, I doubt we’ll be seeing a data centre that an environmentalist would consider to be green for a long time yet.