IBM, Cisco, VMware talk up emerging green tech
At the Business Goes Green conference in San Jose, Calif., on June 6, Forrester Research Senior Vice President Christopher Mines moderated a discussion on data centers and the green technologies that panel members believe will have the most significant impact in the future. The panel included: Elaine Lennox, IBM's vice president of marketing management; Rob Smoot, VMware's data center product marketing manager; and Mike Capuano, Cisco's director of routing and switching.
Christopher Mines: What is the biggest bang for the buck? If you could invent a new technology that your company brought to market that would have the biggest impact in green IT, Rob, what would it be? Where is the biggest opportunity?
Rob Smoot: Well, I was told I shouldn't be a commercial up here, but...
Christopher: Oh, of course! It's virtualization. How could I forget?
Rob: We argue that a large bang for the buck is maybe one of the better initiatives to consolidate. What you have is a starting point. They say that the most efficient IT or data center is one that is not there. Right?
And ultimately, you are not going to deploy a technology, at least in IT, that reduces your service levels, that impacts availability. That is kind of the Holy Grail. You have to get those things, at least maintain it if not improve it, and also become more efficient and more green.
So, server consolidation, because data centers and PC's, all of the IT equipment is so grossly under utilized, five or 10 percent... In fact, I saw a study, McKenzie did a study with the Uptime Institute recently. They found that 30% of servers were only one to three percent utilized; basically dead. Just as one example.
Driving up that utilization to 80 or 90 percent has the same amount of savings. The savings just goes directly to the bottom line in terms of energy. We argue that is the foundation for Green IT. You want to start with IT. When you are looking at a data center you are looking at the cooling side as well, because it has a multiplier effect.
Everything you get rid of on the IT side you then don't have to cool the power generation in all the facilities there. Usually it takes more than double your energy, in even a very efficient data center or IT environment, to cool everything down.
But, beyond that, it is not just about the consolidation. Consolidation is really not a one time thing. It is an ongoing effort. At any given point in time, you should only have the IT equipment running that you need to deliver the computing and the service levels and the availability to the business.
And so, as it results, that is largely where we are focused on now. Once you virtualized, now you can move systems dynamically. The underlying hardware becomes less relevant. You can move a running virtual machine from one physical server to another.
And as a result, you can do some very interesting things with power management where your systems are still running, they are just running on much fewer services and we are automatically powering down the servers. It just constantly is staying on top of itself and right sizing IT at any given point.
Elaine Lennox: You can apply the same logic at the data center level. At IBM, we completely agree with everything you said. Virtualization is absolutely the place to start. But, you can apply the same logic at the entire data center level. Rob was talking about the systems, right? How do you become efficient within the systems? Well, the various studies say that 60% of your power is actually cooling of the data center; the physical place where your systems are sitting, right? Depending on his estimates, 70% or however much, but a big chunk more than is actually used by just the systems.
One of the things we do at IBM is we will go in and do a really quick assessment, very low cost assessment, for clients to actually look at their data center and say "OK. What could you do that is really easy?" I think, the best example is cabling. You wouldn't believe this, but the number of times we can just send a basic technician into the data center, pick up the floor tiles, and there is a mass of cables under the floor tiles.
Well, the whole reason we had raised floors... people have forgotten why we have raised floors in the data centers; so the air could flow under the floor! So, what do we do? We have got this mass of cables under the floors. So, literally, send a tech in, straighten....
Christopher: I thought it was just stash all the old junk under there so it was out of sight.
Elaine: Yeah, stash the old junk and the old cables, right? Send a tech in, straighten up your cables, and you can reduce your power consumption by sometimes as much as 10 or 20 percent. I mean, that is insane. We are talking about just putting rubber bands around your cables. That is not terribly high tech compared to our world, but there is basic stuff like that. That is one of the things we try to do; encourage clients to just get someone in to help them just look at the basics of how they are running their operations.
I was thinking earlier, the other one is hot spots. This is something in a data center, and this is something we all do inherently. Because what we do is we say "OK. Here is the cluster of new servers that we are going to use for this new application. We will put them over there. Here is the cluster of servers that we are going to use for this other thing. We will put them over there. Here is our third application and we will put it over there."
Well, what you do by doing that is instead of taking the servers that are going to be running at peak utilization at 12 o'clock during the day, during your peak work load, and putting one in this corner and one in that corner so they are not creating a hot spot, you just put all of the servers that are going to run at the same time of the day and create all of the energy in one place.
So, you create this hot spot in your data center because it gets this peak work load and it is running like crazy. And of course, you have to cool the entire data center because of that one hot spot, which created too much heat. So, it is really simple. Just move some of the servers around. It is not rocket science.
But, there is really simple stuff that clients can do that we can come in and just do a really quickie. I am not trying to be a commercial here. But, IBM or any other vendor can do a quick assessment for you and say "OK. Well, here is some easy, low cost; doesn't require buying any new stuff." So, it is kind of the same logic. But, again, I agree with you in starting with virtualization on the system side.
Mike Capuano: I mean, virtualization is certainly a low hanging fruit, right? I was going to take a more motherhood and apple pie answer to the question. I think, instead of focusing on low hanging fruit, what is the big impact? I think long term it is about this concept of monitoring, measuring, and managing supply and demand.
We have talked a lot about consumption here so far. I think, the other side... and I don't know if it is in your agenda today, but the Smart Grid and really managing the production of power and balancing co generation from a more sustainable type of power sources like wind and solar.
I mean, if you look at the electrical grids today, they are built on 1950's 1960's technology. They are coming out and reading a little analog spinning meter inside a glass bubble on the side of my house and I work in Silicon Valley.
So, you look at the power outage and the problems in New York City. It is kind of similar to virtualization in a sense that the reason that power outage happened is because there was no communication between the networks. It just rolled through and boom, boom, boom it took down the whole city. I had friends who walked like eight hours to get home to New Jersey.
So, just having packets flowing from the electricity, for example, and being able to understand what is going on and communicate demand. Like in California, we have the rolling brown outs sometimes. If we can intelligently control supply and demand and reduce demand at those peak times, I think, in the longer term that is where we have to get integrating intelligence throughout our infrastructure.
That is just fixed assets. You can also extend it to mobile assets. Theoretically, in a connected city, for example, control traffic flow; control lights to make sure that traffic is flowing as efficiently as possible to reduce emissions. Companies like Fed Ex can use it to manage their energy consumption.
So, you can extend it even longer to, I think it is a little longer term thing, but to mobile assets as well. I think, that is probably a long term, but the biggest bang is that sort of thing.