John Lamb, an IT architect with IBM, is the author of the new book, The Greening of IT, a guidebook for optimizing IT infrastructure from top to bottom. Aimed at any level of the organization, from CEOs or CIOs to data center managers and sysadmins, the book digs in deep to some of the best existing ways for making IT systems as energy efficient as possible.
At the end of our wide-ranging talk, I asked John to walk me through a thought exercise that lays out the green IT projects that make the most sense for three kinds of companies: those just starting out, companies with some experience and upper-level buy-in, and companies that have gathered all the low-hanging fruit. That thought exercise has been posted as a podcast, and the audio and full transcript are online at GreenBiz.com.
Matthew Wheeland: You're currently in South Africa working on an IT project, and I'm sure you travel all over the world doing this work What is your sense of the growth of green IT around the world, in the U.S. and abroad?
 |
John Lamb: Well, actually, it is very interesting that certainly things like green data centers and energy-efficiency is very global, and South Africa, although it's a developing economy, it's certainly much more developed than other parts of Africa. So we're seeing, in working with data centers in Africa, and including South Africa, there is a lot of potential for green data centers, especially for energy efficiency.
We've actually maxed out the power that can be supplied to the main data center here, so the customer -- a cell phone company -- is very interested in that. So it actually has shown me that this is very much a global area, and it's something that we can all contribute to -- just energy efficiency, of course, but even in green IT because we all are users of IT through cell phones PCs and the like.
MW: And it's interesting, just the other day at IBM's Green and Beyond Summit, I had a really interesting conversation with an IT manager for a county in Georgia, and he talked about how the county basically ran out of electricity availability. They'd hit the wall, and we were talking about drivers that he had seen towards these green data center projects and other green IT projects.
So I'm curious, what have you seen as sort of the most powerful driver towards undertaking any kind of green IT project?
JL: Well, actually, in some of these areas running out of power is a great incentive, and it doesn't happen just in data centers in Georgia or in South Africa because -- and I mention this in the book -- Canary Warf in London ran out of power. This is the big financial area, and they were told the power company couldn't supply any more power, and south of 14th Street in Manhattan just hit the limit -- Con Ed could not supply any more power to data centers.
 |
So that is a crucial area, but mostly there is the financial incentive where you can save significant money. If you've got a large data center -- 25,000 square feet or so -- and you're paying 12 cents a kilowatt hour, just through energy efficiency you could cut that by 40 percent to 50 percent, which could be $2.5 million a year. So there are very substantial financial incentives to doing this. So I think that is a big hitter.
We could also go on and say, okay, well, there's corporate responsibility, and we all want to help the environment, but the cost savings helps.
But the nice thing about green IT is that without even considering being the good corporate citizen, just the financial incentives themselves are enough to make everybody, especially the CFOs, think about, "Well, this is a great thing to do, great payback, a good business incentive."
MW: Along those lines, I'm always curious what the surprises are that your clients find in the months after one of these projects has been completed. Is there anything that jumps out at you?
JL: Well, actually, the thing that jumps out right away, and this happens within IBM, too, is that before you can start off on energy efficiency, you have to baseline it. What are we using right now?
I've seen this with customers and within IBM itself; you can't measure it because there's one electric meter for the whole building, and in some cases for customers, too, the data center is only a small part of it, maybe 5 percent of the square feet. So it's very difficult to measure the power that's used.
So the old mantra is "you can't manage it if you can't measure it," so that's one of the things that continues to be surprising because customers, and even within IBM and certain data centers, they don't really know how much energy is being used.
Another area that is really important is to understand [data center energy use] and to have some sort of chargeback. If you don't have some sort of a chargeback -- for example, if the CIO doesn't pay directly the costs of electricity for the data center, if it's just allocated on the square footage basis for the whole building, then that greatly reduces the incentive.
So first thing to do is to measure what the data center is using and probably have some sort of chargeback where you can say, "Okay, Data Center CIO, this is what it's costing, and we want you to pay up on that." And it's amazing, and actually we see it all the time within data centers, too.
If I'm a developer and I want so many gigabytes or terabytes of storage, but if you don't charge me back or tell me what the cost is, as a developer if I think I need 400 gigabytes, I'm going to ask for 800 gigabytes. And I have to be charged back to keep that in check. There has to be some accountability.
MW: And that whole idea of chargebacks is one that has long been cited as one of the big barriers to overcome for IT, is that there is no real accountability for how much energy the data center is using.
I've seen studies recently that suggest that that might be changing, some of which say it's changing quickly, and some say it's changing more slowly. Are you getting a sense on the ground that that silo between facilities and IT is breaking down?
JL: Oh, absolutely! That in order to look at energy efficiency and the like, it's absolutely necessary to measure, and it doesn't mean just putting in the meters. You can do a lot of things within the data center itself to actually measure it. And there are a lot of software tools now that can actually help you not only measure it, but manage it. There is a lot of what they call "agentless" technology where you can just grab what energy is used by your servers, by your storage and the like.
And there's a lot of incentives for that. And as you pointed out, it is one of the areas, even in this down economy, that's getting a lot of incentive to save money with data centers and IT in general because the payback is so good.
MW: I found the book really useful in part because it walks through every step of green IT, both from kind of a top level perspective, but then also on a very detailed level.
One of the things that really stood out for me, which I hadn't seen very often, was your discussion of software's role in this. It seems to me, as someone who looks at this every day, that software operations and optimizing them is one of the most overlooked elements of energy efficiency. So I'm curious about your recommendations in terms of software projects to take up?
JL: Well, actually, it's interesting that you bring that up because software is one of the harder things to measure as to how efficient you are. In fact one of the examples I have, and I'm an IT architect -- not so much the developer anymore -- but we had a big data warehouse that was taking so much time to create the reports that it would take more than the window that they had -- more than 12 hours. And we had the sysadmins getting onto the servers and adding CPUs, which you can dynamically add with virtual servers now, and adding more memory, and we'd still peak out. It would still take all of this time.
But they went back in the data warehouse and said, "If we're gonna change the scripts, rather than looking at the whole huge database every time, it's just going to look at the part it needs." And believe it or not, this is astounding to me, this report that took 12 hours of grinding went down to 12 minutes.
And of course this is not a surprising thing to anybody who's programmed. You can always run a program that's very inefficient. But if you think about it, if it took 12 hours, and it went down to 12 minutes, that 12 hours does mean energy because you're grinding away at the server on that time.
And the other related topic with applications that's hot right now is the deduplication of data.
It is amazing how much storage companies are using now: IBM has projected that in the next 10 years, CPU requirements will go up six times, but data storage will go up 69 times. And when you work with companies it's not surprising because it used to be we were talking about gigabytes; now we're talking about hundreds of terabytes of data, even in South Africa, because they wanna keep records of all of the data.
And deduplication can help this because often you have a tremendous amount of data that's repeated many times. And I can say from my own email, I keep the same documents many, many times. But software exists now that can go through and say, "Okay, I'm not gonna have the same document stored ten times. Just nine of the times I'll just have pointers to the one data." And they've found that you can compress and reduce your data storage requirements by anywhere from one-third to nine-tenths.
So these incentives -- and they are related to energy to a degree, but of course just efficiency in general -- are driven by costs, and so just like in the economy in general, when the economy goes south it has some advantages because people are much more inclined to look at, "How can I make things much more efficient?" including all of their IT.
MW: Looking ahead to the medium or long term, I'm always curious to know what is needed or what's in the works that might move green IT technologies to the next level. Are you seeing any noticeable technologies or trends out there?
JL: Well, most of the low-hanging fruit in green IT are basic things like reducing the number of data centers, reducing the number of servers, and we had a case in India where we worked on 35 data centers that were then reduced to seven. And that's the basic technology. We just consolidate and don't have so much extra resource.
But a lot of it is also a sort of "back to the future," because it used to be that in the mainframe days everything was concentrated in the data center, and then we spread out the technology all over the place and each department had their own servers.
A lot of the consolidation and basic practices are just now coming back; one re-emerging trend that's coming back is water-cooled systems, which a lot of the data center people still hate, but it's so much more efficient.
When you get down to the really intricate technology, even the chip manufacturers, including IBM, have gotten to the extent where if you've got a 10-CPU server, that it can look and say, "I don't really need that; I'm gonna shutdown the power to five of these, or reduce the power." So you can get really sophisticated and granular in that area.
And as we know that all of these other technologies for generating electricity are becoming much more economical -- not only wind, but photovoltaics. It's interesting how all that energy and technology fits together so we can say, okay, green IT, that's one thing; but green IT or IT uses electricity, and so all of the other technological advances that we know about, like alternative energy, comes into play.
MW: And that's interesting, because it's true across the board -- of all the types of business activities that I look at, IT is starting to make headway into all these other areas so that you have IT controlling the building energy management and IT controlling logistics for shipping.
JL: Right, and that's an interesting aspect I was just thinking about, too, because IBM is very much into the smart grid. Because the electrical grid in the U.S. wastes so much power, but with the smart grid, you get feedback from people as they cut down their power use, so utilities don't need to generate all this electricity that will go down the drain and be wasted.
But it's IT that makes the smart grid. You have all this continuous digital information that comes back, and you have to be able to manage it. And it's all related to IT.
So IT is at the core not only of all businesses, but of all these energy efficiency areas coming down the pike.
MW: Given that IT managers and IT departments are facing the same budget constraints that every other department is facing, how do you suggest that they get buy-in for ambitious projects? Or is the solution to move away from ambitious projects and adopt more shorter-ROI, piecemeal projects?
JL: Well, in a way it depends on where you are in the green IT or green data center area, but certainly go for the low-hanging fruit. For instance, if you're in an area where you can use outside air for cooling -- as we know, cooling can take almost half of the energy in a data center, so that is a tremendous opportunity for savings. And it's the simplest thing in the world, but there's a lot more incentive to do that now.
And just migrating the servers -– if you start with the older servers, that will give you the largest savings first. And servers, and I tried to point this out in the book several times, are just like our laptops. Companies refresh the servers every four years or so just to keep technology up, and this provides a great opportunity to get in more efficient servers.
Then there are still plenty of servers around that are very lowly utilized. Those are the ones to consolidate first. Really, just looking at the oldest technology first because they always use much more power is always a good first step.