ms datacenters : economic downturn

Rightsizing Servers to Achieve Cost and Power Savings in the Datacenter

msdcblog — Tue, 15 Dec 2009 15:07:00 GMT

By Dileep Bhandarkar, Ph. D.

Distinguished Engineer

Global Foundation Services, Microsoft

I speak about server rightsizing from time to time at technology conferences, and am surprised that the basic concepts are new to some people in the industry. A lot of people don’t realize the huge performance disparity between the Central Processing Unit (CPU) and disk input/output for instance, and the opportunity that this disparity provides to achieve significant savings by balancing the system.

Rightsizing can clearly reduce the purchase price and the power consumption of a server. But the benefits go beyond the savings in capital expenditure. The lower power consumption has a big impact on the Total Cost of Ownership as shown in the Figure by reducing energy consumption and also reducing its share of the datacenter’s power allocation shown in blue.

Figure: Three-year total cost of ownership of a basic 1U server

It also surprises me that so many IT groups base their purchasing decisions on published benchmark data about processors, even though that data is often generated using system configurations that are completely unrealistic when compared to real-world environments. Most folks sit up and take note when I display the facts about these topics, because the subject is important.

IT departments are strapped for resources these days, and server rightsizing is something every team can do to stretch their budgets. The point of my presentations and the white paper our team is publishing today is two-fold:

1. To quantify some of the opportunities and potential pitfalls as you look for savings, and

2. To present best practices from our experiences at Microsoft, where the group I lead manages server purchases for the large production data centers behind Microsoft’s wide array of online, live and cloud services.

In a nutshell, the paper describes how we perform detailed analysis of our internal workloads and then select balanced servers that are optimally sized for our production scenarios. It is my hope that IT teams in other companies can use the information in our paper to justify devoting resources to characterizing internal workloads, because that is the basis of an effective rightsizing strategy.

Before I came to Microsoft to manage server definition and purchases I worked on the other side of the fence. For 17 years I focused on processor architecture and performance at Digital Equipment Corporation, and then worked for 12 years at Intel, focusing on performance, architecture, and strategic planning. It’s interesting how now that I’m a hardware customer, the word “performance” encompasses cost effectiveness almost as much as it does throughput and response time. As my colleague Kushagra Vaid and I point out in our paper, when you look up performance in the dictionary it is defined as “how well something performs the functions for which it’s intended”.

Today’s IT departments need to address the cost and power aspects of server performance because saving money and energy efficiency are such important goals for companies these days. We hope this paper, and the data and experiences it shares will help you in this regard.

\db

The Polarization of the Data Center Industry (at least for now): Part 2 of 2

msdcblog — Tue, 08 Dec 2009 04:21:00 GMT

The Inflection Point for Efficiency in the Data Center

By Christian Belady,
Principal Infrastructure Architect
Global Foundation Services, Microsoft Corporation

The Gartner graph (included in Part 1 of this blog) also made me think about the fact-- that had the same survey been asked 5 years ago-- it would have shown100 percent are either looking at internal metrics and/or not considering anything when it comes to energy or efficiency. In my opinion, it wasn’t until 2006 that the industry really did go through a paradigm shift. While there were a few of us who had been pushing efficient computing approaches for over a decade), we had limited success in moving the industry until 2006. What happened in 2006? I think the notion of establishing an industry efficiency metric and data center metrics were born. I thought it would be interesting to recap those milestones based on my perspective:

January 31 2006: Conference on Enterprise Servers and Data Centers: Opportunities for Energy Savings

This conference really brought Efficiency on the main stage of our industry. Andrew Fanara’s (EPA) presence was the key to waking up the industry and his goal was to drive metrics for delivering efficiency into computing. This (to me) was a pivotal day since it marked the first time we collectively got together as an industry to address efficiency. It was no longer a voice of a few, thanks to Andrew.

March 27, 2006: Energy Efficiency Server Benchmark Technical Workshop

This event was a follow-up meeting to the EPA conference in January. I led a work group on the Applicability of Benchmarks to the real world” with members from HP, Intel, AMD, LBNL, Cisco, Dell, Sun, Rumsey Engineers and others. The group concluded that a benchmark should be developed with power on the y-axis and workload on the x-axis. This recommendation is what drove the publication of the paper later that year that Jon Koomey spearheaded (with many of us in the industry co-authoring) titled “Server Energy Measurement Protocol”. This paper was the foundation of what SPECpower is today.

April 23-26, 2006: High-Density Computing: Trends, Challenges, Benefits, Costs, and Solutions

This symposium was the Uptime Institute’s first and it focused on Density trends. However, it was this conference where I first presented an Efficiency Metric called PUE which seemed to capture the attention of many of the attendees. As a result, I published a paper on PUE with my good friend Chris Malone later in the year at the Digital Power Forum. At this same conference, AMD’s Larry Vertal and Bruce Shaw sat down with Paul Perez (my former VP) and I to discuss the idea of starting a consortium called the Green Grid. Ten months later the Green Grid was officially announced with one of its first whitepapers evangelizing metrics and in particular PUE.

These three events to me really started it all. If you look at the Gartner graph, you can argue that all of the metrics were seeded in these three events and are really making an impact today.

So our industry woke up in 2006 and while the Gartner graph does show we have work ahead of us, I do think we can say that in less than four years the industry has made great strides (and perhaps I shouldn’t complain so much!).

/cb

For more information on Microsoft's datacenter best practices visit our external web site at www.globalfoundationservices.com and check out Rob Bernard's feature “Using the Power of IT to Drive Environmental Sustainability” that highlights the effort Microsoft is making in environmental sustainability, as well as how that directly relates to what world leaders are currently discussing in Copenhagen for the United Nations Climate Change Conference 2009.

Building a Better Mousetrap a.k.a. Optimizing for Maximum Efficiency in an Economic Downturn

msdcblog — Fri, 23 Jan 2009 18:31:00 GMT

By Arne Josefsberg and Mike Manos, January 23, 2009

As you might have read in Microsoft’s Q2 FY09 earnings release yesterday, the company has announced cost management initiatives due to the global economic downturn. And with the earnings release back in October, Microsoft announced a reduction of projected capital expenditures by $300 million to our data centers. You might be thinking that the data center team is pulling our hair out trying to figure out how to meet our goals given the new constraints. After all, we need to continue supporting a growing base of more than 400 million Hotmail users and over a billion Live Search queries each day, plus 250 other services for Microsoft, including a fast-growing online services business for enterprise companies and the new Azure platform that software developers are beginning to use to create new services.

But we’ve been preparing for lean times for a while. This recession is the ideal backdrop to implement small changes that target big needs. Frugality drives innovation, and limited resources are just another forcing function to develop creative solutions to infrastructure needs. For our industry, this means more reasons to identify the small tweaks to products or operational approaches that can unlock big opportunities.

If you’ve been reading our blogs, you know that a major focus of our team (for a couple of years now) has been on smart growth and making gains in power optimization, server utilization, and efficiencies that will enable us to do more with less. And we have been a proponent of the industry, as a whole, to start measuring PUE in their data center facilities. This provides a great foundation on how to manage and view your operations.

We measure everything in our data center operations to the point of obsession, and continually analyze our measurements to identify areas where we can improve efficiency and increase ROI. Of course we’re doubling down on those efforts in light of the new budgets constraints, but we’ve already been moving in that direction for quite some time. A big part of our efficiency efforts involve increasing the utilization of our servers—an area where our industry is now beginning to raise the bar. We’re working on a multitude of fronts in this area and we are making real progress. One way is through virtualization, where we’re beginning to run multiple service tasks on the same machine. Another is by working closely with our internal customers, providing them real-time data about their utilization to help them identify extra capacity they already have so they don’t need to order more servers. And our research and engineering teams have developed some interesting and innovative approaches to increasing power densities that help us get more capacity out of our existing resources.

Thanks to the efficiencies we’ve gained through these ongoing efforts, we will be able to delay the construction and opening of some of our facilities, which will save Microsoft and its shareholders significant operating expenses, going a long way towards meeting the goals that Microsoft announced this week. For instance we’re postponing construction of the data center in Iowa that we recently purchased land for. We are still continuing construction of our facilities in Chicago and Dublin, and are planning to open them as customer demand warrants. But given the current economic climate we’re going to do the right thing for our business and shareholders and revisit our plans on a quarter-by-quarter basis. On other fronts, we are expanding existing capacity and making improvements for our other co-location facilities, like Amsterdam, that strengthen our global footprint and help us meet growing demand for online services for businesses.

The bottom line is that despite the problems the economy is going through, our online services businesses are growing. We expect that more companies will turn to our services to save money – by allowing them to decrease overhead costs – and that software developers will increasingly use the flexibility and low cost of entry of our new Azure platform.

We have been busy “building a better mouse trap” for these type of scenarios and are now turning up the dial on our efficiency efforts. Thanks to a lot of hard work and innovation by our team in recent years, we are prepared to address market changes – without requiring Microsoft to skip a beat in moving its Online, Cloud, and Live service businesses forward.

Be sure to visit our team blog next week when Christian Belady, our power and cooling architect, will be sharing more information on our efficiency best practices.

aj/mm

Arne Josefsberg, general manager of infrastructure services, Global Foundation Services, Microsoft
Michael Manos, general manager of data centers, Global Foundation Services, Microsoft