/>
X
Innovation

Sun Exec: Your 'Eco Data Center' doesn't have to cost millions. Some tips for getting there.

OK, so here’s the $100 million question: Do you really want to spend that much to make your data center more green or at least more efficient? I thought not, which is why I agreed to speak with Sun Chief Technologist for EcoDC and Virtualization Services, Phillip Morris, about more practical matters.
Written by Heather Clancy, Contributor on

OK, so here’s the $100 million question: Do you really want to spend that much to make your data center more green or at least more efficient? I thought not, which is why I agreed to speak with Sun Chief Technologist for EcoDC and Virtualization Services, Phillip Morris, about more practical matters. Following are excerpts from our discussion about short-term, medium-term and long-term best practices you can adopt to help your own company. And, yes, even changing out the fluorescent light bulbs will help.

GreenTech Pastures: Who owns the data center problem? Answer: That’s a very important question. Dave Douglas, who’s [the Sun] Chief Sustainability Officer, makes a great statement: ‘It is everyone’s responsibility in the company.’ However, that doesn’t tell you who owns it. Everybody has the responsibility for making it happen. But unfortunately, if you look across what impacts and the areas of responsibility that are impacted by doing an eco data center or a green data center, or whatever, you have people like the VP of facilities, who usually reports up to the CFO. Obviously, the IT folks, who report up to the CIO. You have business application guys. You have the business application users. You have the COO and supply chain people, who have a big piece of this for both RFPs, for sourcing and for disposal. So all of those aspects. There is only one single point of focus that drives up to, and that is the CEO, unfortunately. Or possibly the COO, if the company has one. However, the CXO level typically doesn’t want to spend day-to-day time thinking about their data center facilities. …

Sun has a perspective that an ownership should be created that crosses these boundaries, has fingers into all of these different groups but that probably is a small group in and of itself. In fact, at Sun, our Eco Responsibility group only has about six people in it: a VP and then representatives into each of the affected groups. ….

It doesn’t matter where this lives, but it has to be kind of outside the scope of the regular parts of the architecture of decision-making. You can’t have the CIO have too much influence over it or the VP of facilities have too much influence over it. But it does have to have its own separate authority, responsibility and budget.

GTP: When you talk to people outside of Sun, do you have the sense that people have made a decision around this? A: There’s a lot of questions around this. And in fact, one of the services that Sun specifically offers is called data center strategy. The very first part of data center strategy is to take a look at the stakeholders involved, the relationships, their goals and metrics, how their [profit and loss balance sheets] work, whether they are at cross purposes with each other, and help create a vision for where the company can go and how they can help develop a single point of focus for helping things move forward. Quite often, that might be a separate Eco Responsibility team or it might be just a data center team for a one-off type of situation.

Read on for specific tips for realistic short-term, mid-term and long-term goals you can set for your own Eco Data Center project.

GTP: Where should the reader of this blog start? A: Certainly having a green data center strategy or a green data center is going to be the place where most often the place that you have to start. If you don’t have a strategy, you can’t then drive towards having principles, which then help you set up objectives and then out of objectives you have short-term, medium-term and long-term goals and then drive toward that.

If you lack the strategy, the whole rest of the vision falls apart. So a lot of times people are getting started in a situation that I feel is somewhat upside down. They will go through and create a number of goals, and quite often the goals are admirable: to do virtualization and consolidation and lower their power bill, whatever they are trying to do. But often times it’s done somewhat in a vacuum, it’s very difficult and there are a lot of political negotiations that have to happen to do virtualization and to get the business users to buy in and the application developers to buy in. It becomes very difficult if you don’t have that top-down alignment. So, to start, first things first. Create the strategy and then that should drive putting together a targeted team to take on full ownership and responsibility. And then the next thing to get started is you should have a baseline measurement—where you are, how you’re doing, where you’re going, what do you consider an improvement. If you really don’t have a baseline measurement, you can’t know if you’re improving or not, you can’t know what impact you are having. So Sun’s statement is measure for awareness and manage for action. Short-Term Gains GTP: Would that be considered to be part of your short-term objective? A: I would say that this actually precedes short-term objectives. If you don’t have the strategy, it’s very difficult to set objectives in place that align with everything that you’re trying to do. So, knowing what you’re going to measure, how you’re going to measure it, how you want to manage it, what your goals are. From there, you can start looking at short-term objectives.

So, for us, the typical things that we want to concentrate on for both short and medium term are things that really help the customer avoid large-scale capital expenditures ,such as having to build out a whole new data center. Instead, using the baseline measure that are in place, we go through a continuous cycle of assessment and optimization. And a number of things we look for are short-term hits that can help your data center live longer, give you continuous value without having to build out a whole new data center, etc.

No. 1 is to have your data center mechanical and electrical systems examined and assessed for possible improvements and efficiency. Make sure your temperature set points are at the high end of the recommended range, 74 [degrees] to 78 [degrees] American standard. Other things to do are to decommission unused assets in the data center, servers, storage, networking gear, etc. There’s studies out there from IDC, Forrester, Gartner that say typically 8 to 10 percent of the assets in a data center are almost completely non-utilized, but they’re still there in place because the business units are uncomfortable taking them out yet. There’s lots of ways that you can actually decommission equipment, and in fact when Sun went through and did our optimization, we decommissioned 5,000-plus pieces of equipment, servers, storage, networking, etc., across the board. These are easy things.

Enabling power management on all the systems that support it, creating a vendor supply strategy that sets metrics around efficiency, using online power calculators, high-efficiency power supplies, low-power CPUs, lower-powered DIMMs and less memory DIMM sockets total because DIMMs themselves are starting to use up a lot of energy. Another area to look at is improving storage efficiency, having less physical copies, using deduplication, higher density drives, using RAID Fibre / RAID 6 instead of RAID 0 Plus 1. On the infrastructure side, you can calibrate your sensors so that one airconditioner is not blowing hot air while others are trying to overcool the situation, so actually managing the air flow analysis, the cooling analysis. Closing doors and sealing windows, etc., so that cooling humidity isn’t spread to other rooms. Fixing humidity levels and sensors that could be driving higher cooling costs. Removing excessive floor cooling grills that are not needed or in wrong locations. …

Those kind of things are very easy to do in the short term and can you give you a lot of power back that was being wasted on infrastructure support such as PDU losses, transformer losses, cooling losses, cooling losses can be really huge, and that allow you to redirect that power to your IT infrastructure, thus gaining back the utilization that you though you might have run out of.

GTP: Are these things you must do before you look at the long term? A: No, but again if your goal, which most customers’ goals align with, is to avoid big capital expenditures, doing everything you can do on that list of short-term objectives can really radically help you to gain cooling efficiency back, gain power distribution back, allow you to actually put more equipment into your floor space that you thought you might have been running out of, etc. Helping you to avoid a huge expenditure on building a new data center. …

In the medium term, that’s where we start looking at systems, storage and network virtualization, consolidation. One of the things that people start to think of when they consider virtualization and consolidation is that is the big low-hanging fruit in greening your data center. But unfortunately it actually only addresses 10 percent of the total costs in terms of power and cooling and management and possibly only 5 to 15 percent of your CO2 footprint. … So even though it is low-hanging fruit, it is not addressing the big piece of the pie.

On the other side, accelerating technology refreshes, ie, replacing old inefficient servers with newer-generation, smaller-footprint, less power-hungry systems. Improving your cooling infrastructure, networking it and integrating the cooling and humidity systems, using free or dynamic cooling. [Using] retrofit kits for variable frequency drive pumps … air and water side economizers, optimized air flow analysis and management, which could be a visual inspection or via computational flow dynamics. Utilizing localized spot cooling where appropriate, delivering high-voltage circuits to the racks to reduce losses, upgrading to much more efficient UPSes and transformers, reducing the number of conversions along the way and distributing high voltage. Consider moving hot spots to containers, such as Sun’s modular data center, etc. The other part is get on the path towards modularity in the data center to match capacity, ie, your power distribution and management, cooling distribution and management, cabling management, etc. There’s retrofit kits for some equipment to help it remain in place, such as the variable frequency drives that I mentioned or, alternatively, redoing an area in the current data center to start implementing Sun’s pod architecture.

For the long term, match your infrastructure to the application SLA or SLO needs, whatever your service level model or service level objectives are. Don’t implement big Tier 3, Tier 3 Plus or Tier 4 data centers for all application workloads. Target it appropriately. A number of customers are starting to take a look at a single data center that is broken up into multiple rooms. Some of the rooms being Tier 1 Plus, some being Tier 2 and finally the other rooms being Tier 3 Plus. And [they’re] putting only the mission-critical equipment in the Tier 3 Plus areas and all testing, development, quality assurance, SAN box systems, staging and lifecycle management systems into the Tier 2. And then the Tier 1 can be kind of the playground for systems administrators to learn things or people to try out new virtualization technologies, etc.

The other thing is to monitor and reduce overprovisioning via upgrading to a modular infrastructure capability. That is really what Sun’s pod infrastructure is about. We have a link to www.sun.com/futureproofyourdatacenter that specifically talks about matching your data center infrastructure to your IT. Quite often, what you’ve seen in the past is a data center buildout, where they preinstall all the UPSes, all the rack units, all the chillers, so it starts out that just one eighth or one-tenth of the data center is filled with IT equipment. Yet, all the infrastructure is already in and installed and turned on, wasting tremendous amounts of energy. Instead, Sun has architected a methodology where [this infrastructure] can grow in place as your [IT] needs change. We just finished implementing our own infrastructure, which initially is set up for a 9 megawatt data center but can grow in place all the way up to 21 megawatts.

What Green Really Means GTP: Can you address what it will cost to really make your data center green? A: It really depends on how green you want to be. What does green mean to you? Let me answer your last questions as a circular way of getting back to what you asked. So, the last question was how much will it cost?

Basically, every proposal needs to be put through and evaluated and priority given to quick ROI items. In general, if done correctly, going green, ie., ecological, should save the company a lot of green, ie., economic. Sun has modeling capabilities and expertise specific to being able to do customer business case analysis, doing what-if analysis, ROI/TCO modeling, design and simulation of business processes and how they can flow all this through their supply chain processes, their disposal processes, all the way front to back. We basically take a look at things from a Design, Source, Operate and Dispose lifecycle. The biggest hit really in terms of being able to go green is really upfront, in the design phase: So designing your RFP process to support your sourcing, designing IT optimization, designing application modernization. You tune your chain of command and documentation, how you implement things. That’s where the pedal hits the medal in terms of going green.

However, going green is, I would say, more marketing hype than reality for customers. What customers really want to do is save money; period, bar none. If you talk to a CIO, CFO, CEO, they want to have less cost from their data centers for their IT, yet have more productivity out of them. That’s why Sun’s literature never says Green IT or Green Data Center, we call it ecological data center or Eco Data Center. Where Eco can stand for Ecological, Economic and the Ecosystem.

Editorial standards

Related

We will see a completely new type of computer, says AI pioneer Geoff Hinton
artificial-intelligence

We will see a completely new type of computer, says AI pioneer Geoff Hinton

Garmin's new Index BPM is the blood pressure monitor that I've been waiting for
garmin-index-bpm-lifestyle

Garmin's new Index BPM is the blood pressure monitor that I've been waiting for

Watch out for this triple-pronged PayPal phishing and fraud scam
phishing.jpg

Watch out for this triple-pronged PayPal phishing and fraud scam