Showing posts with label adaptive infrastructure. Show all posts
Showing posts with label adaptive infrastructure. Show all posts

Monday, November 10, 2008

Solving IT Energy Use Issues Requires Holistic Approach to Efficiency Planning and Management

Transcript of a BriefingsDirect podcast with HP’s Ian Jagger and Andrew Fisher on the role of energy efficiency in the data center.

Listen to the podcast. Download the podcast. Find it on iTunes/iPod. Learn more. Sponsor: Hewlett-Packard.

Dana Gardner: Hi, this is Dana Gardner, principal analyst at Interarbor Solutions and you’re listening to BriefingsDirect. Today we present a sponsored podcast discussion on the critical and global problem of energy management for IT operations and data centers.

We will take a look at energy demand, supply, costs, and ways to develop a complete management perspective across the entire IT energy equation. The goal is to find innovative means to conservation so that existing facilities don't need to be expanded or replaced.

Good energy management is not as simple as just less hardware or better cooling, but it really requires an enterprise-by-enterprise examination of the "many sins" of energy and resources misuse.

In order to put into practice longer-term benefits, behaviors, and measurements, the whole picture needs to be taken into consideration. The goal, of course, is to promote a low-risk matching of energy supply and cost with the lowest IT energy demand possible.

To help us examine these important topics we’re joined by Ian Jagger. He is the Worldwide Data Center Services marketing manager in Hewlett-Packard's (HP) Technology Solutions Group. Welcome to our podcast, Ian.

Ian Jagger: Thank you, happy to be here.

Gardner: We’re also joined by Andrew Fisher. He is the manager of technology strategy in the Industry Standard Services group at HP. Welcome, Andrew.

Andrew Fisher: Thank you, very much.

Gardner: Let's take a look first at the broad picture of larger trends in this whole energy equation. As I say, it’s not simple. There are a lot of moving parts, and there are a lot of mega tends and external factors involved as well.

I suppose the first thing to look at is capacity. I’d like to direct this to Ian. How critical is the situation now where large enterprises with vast data centers are actually facing an energy crisis?

Jagger: I think it's quite critical Dana. Data centers typically were not designed for the computing loads that are available to us today and they have been caught out. Enterprise customers are having to consider strategically what they need to do with respect to their facilities and their capability to bring enough power to be able to supply the future capacity needs coming from their IT infrastructure.

Gardner: Now, at the most general level, is this a case where there is not enough electricity available or that the growth and demand of electricity is just growing so quickly, or both?

Jagger: I think it's both, and there is also a third level, which is how adequate is the cooling design within the data center itself. So, it is a question of how much power is available, of how much can be drawn into the data center, what is the capacity of the data center, and as I said, how that is cooled.

Gardner: We are also, of course, involving green concerns. There are issues around carbon and pollution, and mandates around these issues. We are also faced with regulatory issues and compliance that are of a separate nature, and many organizations are behaving more like service bureaus, where they have service level agreements.

So there is not too much wiggle room in terms of what needs to be adhered to from compliance and/or service levels. What are the variables that companies need to first start focusing on in order to better execute their management of energy?

Fisher: That's a good question. One of the most important things to understand is how they have allocated power within that data center. There are new capabilities that are going to be coming online in the near future that allow greater control over the power consumption within the data center, so that precious capacity that's so expensive at the data center level can be more accurately allocated and used more effectively.

Gardner: This does vary from region to region, and HP being a global company, perhaps we should also take a look at the fact that in the United States, for example, there are limitations from the grid. The capacity of moving energy, even if it can be generated, is an issue, and in the U.K., apparently in the London area at least, there’s been somewhat of a lockdown in terms of use restrictions around the Olympics.

Ian, perhaps you could fill us in a little bit on some of the regional impacts and how this is supercritical perhaps in some areas more than others.

Jagger: I think you have just got it with the example you have used. It does vary region to region, depending on the capacity of the grid, the ability to distribute it along the grid and how that impacts customers geographically. It's not just about power distribution and generation, but it's also about the nascent situation with respect to compliance.

In Europe, we are now seeing countries, particularly the U.K., who have taken the lead in terms of carbon reduction. Legislation is coming on line, kicking in from 2010, but compliance requirements from 2009, where the top 5,000 companies or so, companies that use a given volume of energy or a value of energy, have to justify that usage in terms of purchasing carbon credits which are set against them.

Each of those companies -- and this includes HP U.K. -- need to establish what the energy usage is and show the roadmap -- how they can reduce that year over year towards the legislation that's in play there. It's only around the corner before that's applied in the U.S. too.

Gardner: Now, we recognize that this is a large problem. Many components -- I have heard the phrase “many sins” -- are involved. I wonder if either of you, or perhaps both, could fill us in a little bit about what are the types of past behaviors, approaches, mentalities, and philosophies about energy that need to be reexamined in order to get closer to where we need to go.

Jagger: I think the contrast among the silos between facilities and real estate and IT are based in the contradiction between cost and availability. You mentioned service levels earlier. From an IT perspective, that’s service level agreements to the business in terms of availability, the uptime of equipment. But, from the real estate perspective, the facility perspective, it's about cost control and CAPEX and OPEX with respect to the facility itself.

They have tended to operate in independent silos, but now the general problem we have, which is overriding both of those departments, is the cost of energy. Typically the cost of energy is now approaching 10 percent of IT budgets and that's significant. It now becomes a common problem for both of these departments to address. If they don't address it themselves then I am sure a CEO or a CFO will help them along that path.

Gardner: How about it, Andy? What sort of sins unfortunately have people overlooked as a result of lower energy cost in the past, but that really can't be overlooked now?

Fisher: First of all, it's a complex system. When you look at the total process of delivering the energy from where it comes in from the utility feed, distributing it throughout the data center with UPS capability or backup power capability, through the actual IT equipment itself, and then finally with the cooling on the back end to remove the heat from the data center, there are a thousand points of opportunity to improve the overall efficiency.

To complicate it even further, there are lot of organizational or behavioral issues that Ian alluded to as well. Different organizations have different priorities in terms of what they are trying to achieve. So, there is rarely a single silver bullet to solve this complex problem.

You need to take a complete end-to-end solution that involves everything from analysis of your operational processes and behavioral issues, how you are configuring your data center, whether you have hot-aisle or cold-aisle configurations, these sorts of things, to trying to optimize the performance or the efficiency of the power delivery, making sure that you are getting the best performance per watt out of your IT equipment itself. Probably most importantly, you need to make sure that your cooling system is tuned and optimized to your real needs.

One of the biggest issues out there is that the industry, by and large, drastically overcools data centers. That reduces their cooling capacity and ends up wasting an incredible amount of money. So we have at HP a wide range of capabilities, including our EYP Mission Critical Facilities Services to help you analyze those operational issues as well as structural ones, and make recommendations, in addition to products that are more efficient as well.

Gardner: You raise a couple of interesting points. It's hard to fix something that you can't measure. What are the basic measurement guidelines for energy use?

I have heard of Defense Council on Integrity and Efficiency (DCIE). There is also a Power Usage Effectiveness (PUE). How does a large organization start to get a handle on this? As you say, or it has been mentioned, it's a siloed problem in the past, now it needs to be tackled head on?

Jagger: You have touched on the principal benchmarks that go through the industry there -- the PUE and the Infrastructure Efficiency Ratio, which is the inverse of the PUE. Put very simply, the PUE would be the total power coming into the data center over the amount of power required for computing purposes. So how efficient is that? How efficient is the data center and service of overall power that is required for computing?

In other words, if you need one kilowatt for computing, and your PUE is two-and-a-half, than you need to be bringing 2.5 kilowatts to the wall to be able to run those computers.

They are not perfect, and there are industry bodies that are looking to drive greater elements of perfection out of this. So for example, PUE is a Green Grid Rating System that is generally used, but Green Grid themselves are looking to migrate through the inverse ratio of the data center infrastructure and efficiency ratio, and use that going forward before they can develop the next level.

The principal problem is that they tend to be snapshots in time and not necessarily a great view of what's actually going on in the data center. But, typically we can get beyond that and look over annualized values of energy usage and then take measurements from that point.

The best way of saving energy is, of course, to turn the computers off in the first place. Underutilized computing is not the greatest way to save energy.

Gardner: That dovetails, of course, with a number of other initiatives we have underway, such as virtualization, application modernization, winnowing out apps that aren't being used very much. Service-oriented architecture (SOA) encourages reuse and making sure that common services are supported efficiently.

There is also data center unification and modernization of hardware. All these things come together and ultimately increase utilization, which then changes the energy equation.

The question is how do we make these things work in concert? How is there some coordination between getting the right mix on energy along with some of these other initiatives? Why don't we start with Ian on that?

Jagger: They feed off each other. If you look at virtualizing the environment, then the facility design or the cooling design for that environment would be different. If you weren't in a virtualized environment, suddenly you are designing something around 15-35 kilowatts per cabinet, as opposed to 10 kilowatts per cabinet. That requires completely different design criteria. You’re using four to eight times the wattage in comparison. That, in turn, requires stricter floor management.

But having gotten that improved design around our floor management, you are then able to look at what improvements can be made from the IT infrastructure side as well. I guess Andy would have some thoughts there.

Fisher: There is a wide range of opportunities. Just the latest generation server technology is something like 325 percent more energy efficient in terms of performance-per-watt than older equipment. So, simply upgrading your single-core servers to the latest quad-core servers can lead to incredible improvements in energy efficiency, especially when combined with other technologies like virtualization.

Gardner: Once these organizations start hitting the wall on energy, it behooves them to look at some of these other initiatives, rather than just saying, “Wow, we need another data center at 10, 20, maybe 100 million dollars.” Is that more the philosophy here -- be smart not big?

Fisher: Absolutely. There is a substantial opportunity to extend the life of your data center, and I recommend that you give HP a call and talk to us here. We have a wide range of things that we can help with.

Ian can talk to the services here in a second, but from a product perspective, we’re bringing to market new capabilities in terms of efficiency of the platforms to help you reduce that total energy consumption of the IT equipment itself. We’re also working on unique ways of reclaiming existing capacity. Instead of having to build another 50 or 100-million-dollar data center, you can live longer in the data center that you have.

Gardner: I suppose one of the fundamental shifts recently with the cost of energy going up considerably is that the return on investment (ROI) equation shifts as well. If I were selling systems I need to know, given the harsh economic climate, that I have a good ROI investment story -- that if you invest $10, you can save $15 over X amount of time. The energy factor now plays a much larger role in that.

Perhaps, Andy, you could tell us a little bit about how the cost of energy, instead of an afterthought, is now a fore thought, when it comes to these -- whether it’s worth these modernization efforts.

Fisher: We look at it both from an OPEX, or your monthly cost of electricity -- and that’s rising rapidly, as the cost of energy goes up -- as well as from a CAPEX perspective, with your investment in your data center.

The first thing is to optimize your CAPEX investment, the money you have already sunk into your data center. You want to make sure that from an investment perspective you don't have to lay out another huge chunk of money to build another data center. So, number one, we want to optimize on the CAPEX side and make sure that you are using what you have most effectively.

But, from an operational cost perspective, it's really about reducing your total energy consumption. You can approach that initially from optimizing the energy use of your IT equipment itself, because that is core to the PUE calculation that we talked about.

If you are able to reduce the number of watts that you need for your IT equipment by buying more energy efficient equipment or by using virtualization and other technologies, then that has a multiplying effect on total energy. You no longer have to deliver power for that wattage that you have eliminated and you don't have to cool the heat that is no longer generated.

Otherwise, there are opportunities… We’ve introduced products that help you optimize your cooling, which typically can be up to 50 percent or more of your total energy budget. So by making sure that you fine tune your cooling to meet your actual demand of your IT you can make substantial reductions on your monthly electric bill.

Gardner: Now, how does the Adaptive Infrastructure relate to this as well? It seems that would also be a factor in some of these equations?

Fisher: We are really talking about the Adaptive Infrastructure in action here. Everything that we are doing across our product delivery, software, and services is really an embodiment of the Adaptive Infrastructure at work in terms of increasing the efficiency of our customers' IT assets and making them more efficient.

Gardner: Let's go back to Ian. It seems that, as with many areas like manufacturing or application development, the history has been that you build it and then you throw it over the wall and someone has to put it into production or build it.

I expect that maybe data centers have had a similar effect when it comes to energy. We set up requirements. We build based on performance requirements. And then, oh, by the way, energy issues come as an afterthought.

Is that true and is that the outmoded method, and are we now, in a sense, building for energy conservation from the get-go? Has it become more of a city- or town-planner mentality, rather than simply an architect approach. What's the mindset shift that's taking place?

Jagger: That's a good question. I think you have to address it at all the levels you talked about. At the company level or the enterprise level, you are absolutely right. That has been the mentality or the approach, we need a data center, and we base it where we are. Nothing else matters. Base it adjacent to us.

Energy costs or supply have not been a consideration. Now they are. That's on the basis that you don't have any other complexities coming at you. But, if you are just looking at the strategy for your data centers in terms of business growth and your capacity, storage, and availability requirements that you have going forward, and you do the math, you can understand the size of the data center you need and how that works with respect to virtualization strategies and so on.

On top of that, we have the latest complexities, where you simply don't have the forward view on things. In just the last few days we’ve seen, for example, Wells Fargo buying Wachovia. I’m not sure how many data centers are within those two organizations, but you can bet they are in the scores. Suddenly, we have real estate and IT managers who are scratching their heads thinking, “How on earth do we bring all this together. There are different approaches now being taken at the enterprise level.

At the architects’ level, it would be irresponsible for an architect today not to build energy efficiency into a green field building or any building, not just a data center. It’s pretty much been established that it just makes sense if you are designing a new building to be building energy efficiency into it, because your operating costs will far outweigh the capital expenditure on those building rather quickly.

I’m not sure how a company like HP can influence at the planning level, but where we can influence is at the industry level and at the governmental level. We have experts within the company who sit on think tanks and governance boards. We advise bodies like the EPA. We sit with the leading organizations in energy building design, and discuss how governance with respect to green building design can be built and can be moved forward within the market.

That's how we can start to influence at the industry level in terms of having industry standards created, because if the industry doesn't create them itself, then governmental bodies will do it.

Gardner: It also seems that because it's so difficult to predict all the variables, that a need for modularity has emerged in the data center design, so that the end result can be amended and adjusted without all the other parts being interconnected and brittle. It’s similar to software, where you would want to have modularity in software, so you gain flexibility and it’s not too brittle. Can you explain more deeply how that relates to best energy management practice?

Jagger: The approach that we at HP are now taking is to move toward a new model, which we called the Hybrid Tiered Strategy, with respect to the data center. In other words, it’s a modular design, and you mix tiers according to need.

What has gone on in the past and today is that as an enterprise you may have a requirement for a Tier 4 level of structure, with respect to the data center, which is putting out at 100 watts per square foot, for example. Let’s say, for the sake of argument, that's a 100,000-square-foot data center, but you don't need all that data center infrastructure at a Tier 4 topology.

If you look at how you’re going to structure your virtualization program, you may only need 50 percent of it at Tier 4 for high density computing, and the rest of it can be at a Tier 2 level.

If that were the situation, you would be clearing roughly 25 percent of your capital costs on building that data center. Just doing simple math, if you are looking at 100,000 square feet, that's in the region of $40 to $50 million. So, there are some clear consequences of moving to a hybrid tiered or a modular model.

Gardner: Are there some examples out there that you can give us? It would be great if you could name some companies, or at least give us use-case scenarios where organizations have adjusted, adopted some of these practices, implemented some of these standards, used common measurement practices, and have resisted having to spend $40 million on CAPEX, but also perhaps utilizing their existing resources even better.

Jagger: I think HP is the biggest example. We are the biggest example of designing modularity into our own data centers.

Beyond HP, you could look at supercomputing centers, high density computing -- the Internet service providers, the Googles of this world, and Microsoft themselves. The companies that require high-level resolve, high density and supercomputing typically are moving in this direction. We are pioneering this with our in-house capabilities. We are at the leading edge of this level of innovation.

Gardner: Let's take a look forward a little bit. What can we expect? Obviously, this makes more sense over time. Green issues are going to become more prevalent. Carbon is going to become more regulated. Costs are going to become prohibitive for waste, and the amount of data moving around increases all the time.

Perhaps you can explain the roadmap, the future, some of the concepts around optimizing data centers -- without pre-announcing things, but at least, give us a sense of what's coming.

Fisher: How about if I talk to that one first. One thing that was just announced is relevant to what Ian was just talking about. We announced recently the HP Performance-Optimized Data Center (POD), which is our container strategy for small data centers that can be deployed incrementally.

This is another choice that's available for customers. Some of the folks who are looking at it first are the big scale-out infrastructure Web-service companies and so forth. The idea here is you take one of these 40-foot shipping containers that you see on container ships all over the place and you retrofit it into a mini data center.

In the HP implementation, it's a very simple kind of layout. You just have a single row of 50 U racks. I believe there’s something like 22 of them in this 40-foot container. There’s a single hot aisle and a single cool aisle, with overhead cooling that takes the exhaust hot air from the back and cools it and delivers it to the front.

Using the HP POD you can install any standard equipment into the 19-inch racks and build out a very efficient data center that has a very low PUE or a leading PUE, from a cooling perspective. So that's yet another option in the HP side.

From the product side of HP here, one of the biggest things we’re seeing is that power and cooling capacity is allocated by facilities in a very conservative manner. It's hard to understand exactly how much energy is required for each individual server or blade enclosure. So, there’s typically quite a bit of a conservative reserve that is allocated on top of what's probably actually being consumed.

In fact, if it's in the purview of the facilities team to allocate that power, they would treat it as any piece of electrical equipment and they would just look at what the max power rating or requirement is for the piece of equipment. What we’re seeing is that this can actually overstate the power requirement by up to three times what is actually needed.

So, there’s an incredible opportunity to reclaim that reserve capacity, put it to good use, and continue to deploy new servers into your data center, without having to break ground on a new data center.

Very soon, you’re going to be hearing some exciting news from HP about how we’re going to provide the opportunity for fine-tuned control of exactly how much power the servers in the IT racks are going to actually use.

Gardner: So, not only are we moving toward modularity at a number of levels, we’re bringing more intelligence to bear on the problem?

Fisher: Yes. A key to addressing this problem is to have accurate measurement and the ability to have predictability and control of the actual power consumption of the core IT equipment that the whole infrastructure is supporting.

Gardner: Alright. How about a roadmap, from a strategic point of view, of methodologies and best practices. Ian, what new innovations can we expect along those lines?

Jagger: In all this complexity, it's a relatively simple path to follow. It all starts with discovery -- where are we today? Given what we know about business direction, where do we need to get to? What do we need to be capable of from a business technology perspective that incorporates a facility as a holistic or a hybrid view of those departments combined? What is it that they need to produce to support the business going forward?

Then, you have a gap. The next question is how do we fill that gap, how do we get there? Various strategies can accrue from that, depending on what your needs are.

We would look at that with customers, and we would sit down with them and ask them some pretty basic questions. Do you need to be where you are today? If you are in Phoenix, does the data center needs to be in Phoenix or could it be in Washington state? It’s cooler and you don't therefore have the energy costs that you would in Phoenix. So, let's have a look at that.

What is your position from a corporate social-responsibility perspective with respect to the environment? How visible are you in addressing that in comparison to your industry peers? What are the pressures on you to do that? So, let's have a look at alternate energy sources with respect to your data center.

For example, we have just announced our San Diego facility, which is now powered by solar panels. We are involved quite heavily right now in Iceland, providing geothermal technologies for data centers. So, a question there would be, can you be in Iceland? One issue there would be the question of latency. There are several questions that you would ask in terms of direction and how to get there.

Having answered those, you would move into planning and design phases and we would address those at that point too. We would build into the operation of any given new sites, or retrofitted site, the processes with respect to service management across the facility and IT structures. Service management is now not only about IT, but it’s about the facility as well, and how that is brought together in one motion.

So, it's pretty much a simple lifecycle approach within a complex field, and that will get you there. Along the process, we would be able to give the orders of magnitude of cost and typical ROI based on the strategies that you are looking to undertake.

Gardner: It certainly sounds like being efficient and getting this larger management capability over energy and facilities and resources is becoming a core competency and not an option. Is that fair to say?

Jagger: Yes. I think the spin on that is, going back to the example I just used of Wells Fargo and Wachovia, who do you turn to who can help you with that? You don't face that every day of your life, either within facilities or within IT, and you need help. You need to reach out for where the help is.

Traditionally, in our industry, as we have been discussing, it has tended to be siloed into real estate and into IT. What’s now required is the holistic view of infrastructure. I mean the physical infrastructure and the IT infrastructure. Customers need to reach out to firms that they feel comfortable reaching out to.

I think it was Andy who actually conducted this survey -- so correct me if I’m wrong, Andy. We recently undertook a survey in each of our worldwide regions, all enterprise customers. The finding was that the more the customers themselves had issues that they needed to address with respect to the environment and energy the more likely it was that they were going to come to HP as their vendor of choice.

Fisher: That's correct.

Gardner: Well, clearly if you don't have the holistic view you are going to have to learn how to get one, right?

Fisher: Right.

Gardner: Ian, let me direct this to you. I suppose there is some thought around environmental benefits and green IT, in which people believe that this is an additional cost or an expense. It seems to me, though, from what we have been discussing, that moving towards good environmental practices is actually moving towards good energy management practices too.

Jagger: That's absolutely right. It is not a choice of one or the other. Now, the business outcomes that come from energy management are also environmental outcomes, but there are apparent barriers to implementing environmental solutions, which, as you just, said are actually energy management solutions. Primarily they revolve around the lack of identifiable ROI or the payback period around any green improvement and then the measurements of that improvement itself.

More recently, we’ve been able to show customers the typical examples of how they can move through that environmental curve or that energy management curve going back to the industry standard benchmarks of PUE.

By showing them what a rough order of magnitude cost would be to move them grade by grade through the ranking system of energy efficiency, we show them what that cost would be, what the return would be, as a result of that in terms of carbon savings, in terms of dollar savings, and what the payback period would be based again on those dollar savings.

So, we can have a very strategic, yet tactical, view on how to approach this. A customer can take a larger view in terms of how far they want to go with their environmental approach and balance that with their energy-management approach.

There is obviously a curve here. The larger the investment in improving energy management, then the greater the return. At some point, that return slows down, because of the amounts of actual investment you have put in. So, there is a curve there, and we can show you how to get to any point along that curve.

Gardner: Excellent! We have been discussing the large global problem around energy management and how it has become more critical for IT operations -- energy not as an afterthought, but really the forethought and an overriding stratagem for how to conduct business in IT.

I want to thank our guests today. We have been joined by Ian Jagger. He is the Worldwide Data Center Services marketing manager in HP's Technology Solutions Group. Appreciate your input Ian.

Jagger: You’re very welcome, Dana, I am happy to have taken part.

Gardner: Andrew Fisher, the manager of technology strategy in the Industry Standard Services group at HP. Thank you, Andy.

Fisher: You are welcome.

Gardner: This is Dana Gardner, principal analyst at Interarbor Solutions. You’ve been listening to a sponsored BriefingsDirect podcast. Thanks for listening and come back next time.

For more information on energy-efficiency in the data center, read the whitepaper.

For more information about HP Energy Efficiency Services.

For more information on HP Thermal Logic technology.

For more information on HP Adaptive Infrastructure.

Listen to the podcast. Download the podcast. Find it on iTunes/iPod. Learn more. Sponsor: Hewlett-Packard.

Transcript of a BriefingsDirect podcast with HP’s Ian Jagger and Andrew Fisher on the role of energy efficiency in the data center. Copyright Interarbor Solutions, LLC, 2005-2008. All rights reserved.