I’ve uploaded my APTA presentation about building a sustainable data center to slideshare if anyone is interested. Clearly NCTD is a pioneer in the transit industry, even if building a sustainable data center is cliche in other industries. Not one participant in the room had added sustainability as a design criteria when building their data centers. We’ll see if I made any impression with the community on this issue.
by Angela Miller
On Friday we kicked off the data center project at NCTD, and rarely have I been so excited about the probable success of a project. Our vendors are Roel Construction (Rob Netzer) and Logicalis (Bob Mobach). We were lucky to find vendors that have such competency and experience in data center design, and specifically in the requirements for attaining LEED certification.
This is a large undertaking for a small entity like NCTD. While I can justify the project purely on the long-term anticipated Return-on-Investment, the deal was sweetened by meeting all of the requirements for the Federal Stimulus program.
The overall project has several elements that will hopefully qualify it for LEED certification, including:
In addition, as I’ve mentioned in previous posts, the District has been committed over the last two years to slowly introducing greener, more sustainable approaches to our information technology infrastructure, including virtualization and consolidation, switching to blade server technology, replacing older equipment with more energy-efficient equipment, and testing desktop virtualization.
All of these steps make solid financial sense for the District – an especially important consideration given the tight financial times we are currently facing. I could not in good conscience recommend these investments if I could not show solid ROI for our troubles. While we want to be an agency with a priority on sustainability as part of our mission, it is logical that we could not choose to make these investments if they required a corresponding reduction in service or positions at the District. Instead, choosing to follow the more sustainable path will realize direct operating cost savings on a monthly basis for the District.
We of course face some challenges to our ambitious timeline. One of those challenges is the delay of a system migration project by 90 days after its anticipated completion. This project is attempting to move our Prime System applications (installed in 1986) from a minicomputer to a client-server web interface. We cannot risk downtime on this environment and therefore out-of-the gate may see a 90-day delay. But every IT project faces challenges, and I am confident the team will find an approach that gives us what we need.
Over the next several weeks I will be blogging about the adventures of upgrading our data center. Hopefully in March I will be able to say that we’ve completed the work and that we are on track with our ROI.
by Angela Miller
Today we inked the deal that was two-years in the making: we are officially installing solar panels at our administrative offices. The intent is to at least offset the estimated power demands of the data center over the next 10 years of anticipated growth. In speaking with various solar vendors, and with our data center installation firm, we feel this is an achievable goal.
With the incentives still offered at the federal and state level, and working with our local utility company SDG&E, the business case for solar is a viable one. In a time when my agency – like so many across the State of California – is facing dramatic pressures on our operating budget, any capital investment that potentially decreases monthly operating expenses deserves a serious look.
We have two phases of this plan – we’ll install solar panels on the roof of the building, which turns out to be an ideal location. Not every building meets our criteria – we have significant open space, our roof was repaired recently with a material that is reflective, and our utility sheds are relatively short. These design elements make our roof highly suitable for a solar installation.
The second phase of our project will install carport solar panels in our public parking lot. Our existing lot was designed at a time when there were fewer requirements for heat dissipation. Our black asphalt parking lot gets quite hot during most months, and installation of solar panels will mitigate some of this problem. We are also making good use of otherwise non-revenue generating space.
This project is a stimulus project through the American Recovery and Reinvestment Act. There is significant pressure on governmental entities to use this money responsibly, transparently, and according to the terms of the intent of the act. I believe this project meets all of the objectives of the program: it is a shovel-ready construction project, a long-term capital improvement, beneficially changes the operations of public transit in the community, installs technology that decreases operating costs, provides jobs during the project, invests in green technology, and completes in a timely fashion to stimulate the local economy.
While the solar projects added cost to our data center redesign project, the investment had a solid business case, with a six-year payback period. An additional consideration was the increasing carbon footprint of the technology resources at the District. This investment not only pays off from a fiscal perspective, but also an environmental one.
My message to IT people is to think bigger than our data centers. We have the opportunity to improve our company’s bottom line not only through our core IT investments, but also in how we choose to construct and improve the buildings that house that technology. Work with your facility managers to become more creative in solving data center problems.
by Angela Miller
I tend to assume that IT people (like me) have a fabulous sense of hyperbole. We toss around words like “disaster” and “outage” and “negative business impacts” effortlessly. And many IT Managers panic when we are less than say 97% perfect. We have a sense of the dramatic.
So I figure people might take umbrage with my characterization of my data center ‘blower problems.’ Perhaps you think I am again exaggerating. The picture on the left should correct this.
When I say blowers in my data center, I mean just that – a line of 6 vents resembling aircraft engines pointed different directions in the room. People who enter this room do so at their own risk, especially when the blowers are churning at full speed. Smaller people on the team probably should have carabiners and safety lines.
Since my arrival at NCTD, this device has been one of the more discussed items on the team. All kidding aside, this is not the right solution for a data center. Admitedly, we were in a crisis mode when this modification was made. And nothing against the people who chose to install these, because frankly we would be in far worse shape had they not done so. This installation bought us a year … maybe two.
In my mind, this is a good example of the dilemma facing many SMB entities in their data centers — how do we evolve to deal with our ever-increasing energy, cooling, and storage demands? How do we make upgrades to our facilities in a way that meets the burgeoning needs? And should sustainability be a consideration in our design choices.
One thing is certain – choosing the wrong solution is far from sustainable and misses our return on investment. In our case, we are now facing not only cooling problems since the air flow with this device does not address the hot spot problems in the room, but also humidity issues, and power issues since we’ve used almost all the rest of the power in the room on these air handling units. On the plus side, if I installed a wind turbine in the room I could probably dramatically improve my Power Utilization Effectiveness.
Over the next several weeks we’ll look at some of the options we’ll have to choose from in implementing a different approach to the cooling needs. Will we be able to choose a green option given our budget constraints? Stay tuned…
by Angela Miller
The Internet and press are atwitter this week with the announcement of AMD’s new Barcelona quad-core chip. The chip, also known as the Opteron 64, delivers something the competitors did not: a native quad-core design that allows for sophisticated power management. According to the testing I’ve reviewed, the chip delivers up to twice the performance of the duo-core processors but uses the same amount of power.
The design element which differentiates this chip is its native quad-core design which allows each core to be utilized and managed independently. This is a strong design element from the power-management perspective: in the duo-core paired design, the paired cores generally run at the same power level. So if one core is at 75% power so is the other no matter what the processing requirement. In the native quad-core design, the power requirements of each core are managed independently. This simple design change delivers significant power savings.
While some reviewers are saying that AMD is very late to the quad-core game, I believe their design philosophy and the significant power savings prove worth the wait. In addition to the native power savings this chip provides, the sophisticated tools for server virtualization are very strong. Strong enough that Rackspace Managed Hosting decided to deploy the chip after rigorous testing throughout their hosted data center.
We should see over the next several weeks testing centers putting this chip through the paces versus other competitive offerings. I look forward to seeing what the guys at Tom’s Hardware have to say toward validating the performance statements from AMD’s marketing department.
I relied on the following sites for analysis in support of this post:
by Angela Miller
With the focus ever increasing on energy efficiency in the data center, the topic of virtualization takes center stage for many organizations. Both the performance of the newly-launched VMWare stock on the New York Stock Exchange (from around $29 at IPO to almost $68 today) and Microsoft’s impending push into the virtualization arena demonstrate that virtualization will be a hot topic over the next year.
But the benefits of virtualization are not limited to large data centers, and many organizations miss out on the energy efficiency and cooling benefits of deploying virtualization even in their small IT shop.
According to Gartner, server virtualization is just entering the peak of the hype cycle and it is 2-5 years from “mainstream adoption.” This means that over the next couple of years, IT managers will watch as over 50 vendors vie for marketshare in this re-energized space.
Ironically, virtualization is a mature discipline with over 30 years of history in the IT world. It was a key component of any mainframe deployment as early as the 1970s. But with the proliferation of servers and storage devices and burgeoning IT data centers, virtualization has come around again as a way to more effectively utilize the IT resource.
The impact of virtualization on the environment may not be immediately obvious. But think of it this way: in a company where the trend is for users to specify they must have servers dedicated to their function, virtualization is a way to segment pieces of the IT infrastructure to provide those dedicated services without requiring the purchase of separate machines. One powerful server can be segmented into dedicated, non-overlapping functions.
This is a powerful utility, especially when you consider that according to various studies the average usage on a dedicated, stand-alone, functional server is less than 10%.
Virtualization comes in many flavors, from server virtualization through hardware and software, to storage virtualization, to the ability to combine many machines into one virtual device.
Server virtualization is either hardware, software, or processor-based. In a hardware scenario, software like VMWare is installed at an operating system layer and allows emulation. Emulation simply means that a server running Linux for example could also run Microsoft Windows inside of a window within Linux. The basic idea behind this virtualization is that the intermediate software intercepts calls from the emulation window and translates those calls into the underlying operating system to pass to the hardware. The virtualization tool simply serves as a translator between the different environments.
There are many reasons to do this, even at a desktop level. For instance, if you are a small business and you have chosen Apple as your platform, you can install tools to allow you to still run Microsoft Windows within a window on the Apple platform. A small business might do this when they are required to use a particular Windows-based tool to interact with a client and do not wish to procure a new machine. From an environmental perspective, emulation allows companies large and small to run multiple operating systems without requiring acquisition of additional hardware.
The second approach to virtualization is to segment the device into multiple operating systems and allowing these environments to operate independently and simultaneously on one device. In this scenario, the server would run both Linux and Windows at the same time and there would be no conversation between the operating environments – each would make calls to the hardware separately and natively. This approach grew more popular with the release of blade type servers by companies like IBM, HP, Dell and Sun.
Blade servers are quite effective in the data center from an environmental perspective. They are typically stripped of redundant components like power supplies. They are compact and have a small impact on the data center physical footprint. And they are efficient because many servers share so many components. Less waste, less energy, less real estate … all good things from an environmental perspective. (One potential downfall of a densely-populated blade server environment is cooling cost. We’ll cover this in a separate post.)
At the component level, some processors by Intel and AMD now come with native functionality to support virtualization. With newer multi-core, multi-threaded processors, the processor can deliver on the virtualization promise by increasing performance by 3-5 times within one processor (according to IDC). Given the heat generation of the processor, the ability to minimize the number of required processors is highly desirable from an environmental perspective.
Server virtualization clearly delivers significant energy savings to companies that can consolidate their environment effectively.
Like server virtualization, storage virtualization comes in many flavors and tends to be far more complex. The primary theories on storage virtualization are appliance-based, switch-based, and storage-controller based. All three approaches basically rely on mapping tables to route application requests between a virtual location and a physical location on the disk. Without detailing the specifics in each scenario, the bottom line on storage virtualization is that it allows large data centers the ability to deploy very large, disk-dense devices that again are far more efficient than a distributed set of independent devices.
So to build on the blade server discussion: deploying many stand-alone servers each with their own disk storage that is minimally utilized is far less efficient than installing blade servers without any native storage and then linking them to a large storage array or network as a separate, optimized device.
From an environmental perspective, one can argue which architectural approach to storage virtualization is more efficient, but the use of storage virtualization in any context is far more energy, cooling, and footprint efficient for the enterprise data center.
Server and storage virtualization should be key components to any green IT strategy. Both deliver significant energy and cooling cost savings to the enterprise and SMB through consolidation of resources.
I relied on the following sites for analysis in support of this post:
PodTech posted an interesting discussion with HP’s Senior Vice President of Technology Services. While most of the big IT vendors are making a play in this arena now, this interview was a quick and accessible discussion of one vendor’s approach to power management.
by Angela MillerI read with interest Western Digital’s announcement of their new GreenPower Drives. Their marketing statement is that this line of drives can save a significant amount of energy over all their competitors’ drives. Western Digital intends to deploy the GreenPower technology primarily on their desktop and notebook drives – most notably on the 3.5″ Caviar SATA drives and the 2.5″ Scorpio mobile drives. They will launch their enterprise GP drives in the later part of 2007. Their specifications and cited industry research assert that they can save 4-5 watts over competitor drives in idle mode. I tried to substantiate this number with some simple research of specifications offered by the different main competitors in the space using the 3.5″ Caviar SATA 500GB drive as the primary comparison. Because Western Digital has not yet announced specifications for the enterprise, I chose three primary competitors based on Gartner estimates of the share of the desktop market: Seagate, Hitachi Global Storage Technologies, and Samsung.
| Manufacturer/Drive | Idle | EfficiencyIndex |
| Western Digital Caviar SATA 500GB 7200 rpm | 4 watts | .0006 w/GB |
Looking at these numbers is a little like looking at the MPG estimate stickers on cars so it also seemed prudent to see whether any respected technical testing entities had put these drives through their paces. While a review of Tom’s Hardware, AnandTech, ComputerWorld, Storage IO, and The Tech Report, showed that none of these entities had yet tested the Western Digital GP specification, most other specifications came close to the advertised numbers.From an enterprise IT perspective, choosing a drive that natively uses significantly less power than comparable drives will result in energy savings. If a company converts to drives that will use ½ of the power of other drives, the company might save as much as $10/year/drive in energy cost.But as a former IT manager, I would have a few concerns:- In my experience, IT Managers at the enterprise level make decisions about desktop equipment based on the computer vendor rather than components. Rarely would the IT manager specify a particular hard drive, or make a decision on the vendor based solely on the hard drive they offer.- At the enterprise level, IT Managers depend on vendors to provide an overall solution: you could not approach HDS and request Western Digital as a drive vendor when their OEM providers are HGST and Seagate.- None of the drives will attain peak energy efficiency performance numbers without serious consideration to configuration, software setup, and consolidation. The bottom line- one 1TB drive will use less energy than 4 250GB drives.The important take-away from the Western Digital announcement from my perspective: Western Digital is setting the bar higher for energy efficiency and will force their competitors to rise to that expectation in order to retain market share. This energy efficiency innovation will stimulate the hard drive market toward more future improvements.
I relied on the following sites for analysis in support of this post:GartnerWestern DigitalSamsungHitachi Global Storage TechnologiesSeagateTom’s HardwareAnandTechStorageIOThe Tech Report
by Angela Miller
This week the ecology.IT blog officially moved to a web hosting service that is 100% solar-powered. GreenestHost.com was featured in Ted Samson’s blog 2 weeks ago and officially came online on August 6, 2007.
A couple of things impressed me about this company:
While this is not an endorsement or recommendation that everyone should switch to Greenesthost.com as a service provider, it does highlight that there is business sense in establishing and offering green computing technologies.
by Angela MillerI’ve been reading the US Evironmental Protection Agency’s (US EPA) latest report to Congress on Server and Data Center Efficiency released earlier this week. This comprehensive analysis of the current trends in energy usage and potential future demands for servers and data centers in the United States projects that in the next five years the demand could double from an estimated 60 billion Kwh/yr to over 120. In five years.
In spite of all the other science that precedes it, this study will likely be the impetus for the US to consider stricter rules on energy efficiency for information technology devices. The legislation passed yesterday regarding renewable electricity and new taxes on oil heralds other changes likely to come from the US Congress. Regardless of the whims of Washington, the EPA report confirms the need for companies large and small to substantively change IT strategies.Consider the report’s discussion of power density issues:
Increasing power density can lead to a situation in which companies are forced to build new data centers not because they are running out of floor space but because they need power and cooling beyond what can be provided in their existing data centers. This situation has driven much of the recent interest in energy-efficiency improvements for data centers. If the power consumed (and resulting heat generated) in data centers can be reduced through energy-efficiency measures, the existing infrastructure can continue to meet cooling and power needs, and costly investments in new data centers can be deferred. (USEPA, 2007)
The report highlights simple changes that IT Departments could implement today without significant investment beyond operational changes, many of which echo strategies discussed in previous ecology.IT posts:
In their estimate, these simple operational changes could result in a 20% energy savings relative to current trends. A 20% change right to the company’s bottom line with primarily operational strategies.Over the next few weeks we will investigate some of the proposed “best-practice†and “state-of-the-art†changes proposed in the EPA study. They assert that by implementing these strategies we could save up to 55% of the current energy forecast.
I relied on the following sites for analysis in support of this post:US EPA Final Report on Server and Data Center Energy EfficiencyNew York Times Article on Energy Bill