Archive for October, 2010

Opening and Closing Remarks at the 2010 Data Center Efficiency Summit

Monday, October 25th, 2010

As I recently posted, the SVLG Data Center Efficiency Summit on Oct 14th, hosted at Brocade in San Jose was an excellent event. Having co-chaired the demonstration program and putting on the summit, I may be biased, but ask any of the 500+ attendees and I think you’ll hear fantastic reviews of the event, with all presentations by end-users of their case studies or research projects. There were no commercial presentations. We had a VC panel, IT sessions, cooling sessions, air-economization, detailed discussion of the lack of contamination and corrosion from outside air, containment, containerized data centers, new metrics and standards for data centers (LEED, Title 24, Revised PUE), etc.

Some folks have asked for me to post my opening and closing remarks, which I am doing here. These are my opening remarks to the full audience at Brocade:

“I have been so fortunate to co-chair this program for a second year. A program based upon a single idea that if we collaborate, we all help each other run and operate our data centers more efficiently. In this forum we educate each other from our own experiences on how we have reduced costs and energy use. Following my own mantra, saving energy is always saving money; this is especially true in a data center.

Reducing energy use has the most economic impact and net carbon reduction–a NegaWatt as Amory Lovins has termed it–for a kWh saved is more efficient and at lower cost than the cleanest kWh generated.

While we can locate data centers in Iceland, Northern Canada Reno, NV with 100% renewable energy supply, the location of our data centers is usually driven by many requirements. While you’ll see case studies today of data centers working to create zero carbon data centers, companies have a fiduciary responsibility to control costs and reducing energy use provides the greatest cost savings and environmental impact reductions for our growing user needs.

With an estimated 17 million servers in the US in 2009 (IDC), every month these servers support nearly 2 billion search users, over 500 million shopping and approaching a billion social networking users; 100 million daily tweets, nearly all of our financial transactions, all of our blogging, and becoming the source for news, knowledge, travel, gaming, entertainment, and sharing and archiving our precious family photos.

The estimated 2.6 million data centers in the US are the hubs of our Internet, and Internet traffic grew a whopping 62% in 2010 after growing 74% in 2009. International Internet Bandwidth has grown over 10 times since 2002. The majority of the growth is from India, the Middle East and Eastern Europe with each growing over 100% per year. And we’re just beginning to accommodate cloud and mobile computing.

While our industry has shown staggering growth, we have become more efficient in what I call the efficiency renaissance of the last few years. We are providing these services at far greater efficiency than methods used only a decade ago. Amazon reports that they can now accommodate twice the servers on the same power draw of before. A recent EPA report shows average US PUE is about 2–which is half the infrastructure usage of average world PUEs of about 3. Google has reported a data center fleet average PUE of 1.2 and I’ve been so fortunate to work on several data centers designs over the last year with annual average energy PUEs below 1.1.

While our best designs will reduce infrastructure energy to 5% of the server load, you’ll see excellent results today with existing data center case studies from simple changes in temperature and humidity settings and air containment. As our infrastructure energy uses a tiny fraction of the energy of our hardware, our focus will shift to reducing the energy use of our IT equipment, as you’ll see today in case studies of software and other IT efficiency projects.

EL Insights reports that our efficiency renaissance is expected to increase the US green data center market from about $4 billion today to nearly $14 billion in 2015. You’ll hear examples during our VC panel of what will lead future energy savings with new processors, networking and memory hardware, as well as new storage, software and lighting technologies.

We have 18 end-user presented efficiency projects and several detailed studies and discussions on various subjects to help you further improve your company’s data center efficiency. I hope you’ll pay close attention to this shared knowledge and experiences of our industry experts and participate in the question and answer periods following each session. If we continue to collaborate, innovate and further improve the energy efficiency of our data centers, we directly improve the efficiency of the Internet and the world’s economy. This will help us avoid energy challenges of the future.

We don’t need to look far back to see the value of our efficiency efforts. For it was in 2001 that the California Energy Crisis hit, and Exodus Communications my employer and the creator of the leasable data center industry was being accused as the cause of the crisis. Our energy efficiency achievements trumped these inaccurate accusations. I also worked with the SVLG to resurrect the Energy Committee that was the original founding purpose of the SVLG in the 1970’s by David Packard of HP fame to address our State’s energy challenges. My peers and I led voluntary energy curtailments to avoid countless more rolling blackouts, and Carl Guardino and I led a group of California elected leaders along with the Governor and the CEOs of nearly every major Silicon Valley company to address these energy issues. Just as we tackled these energy issues then, it is again energy efficiency leading the way with Silicon Valley companies driving the innovation of this efficiency renaissance and the SVLG driving the policy. Through this summit, Carl and the SVLG are part of this continuing effort to advance the efficiency of our Internet industries. Please help me provide Carl a warm welcome as the CEO of the Silicon Valley Leadership Group.”

Here are my closing remarks, which I am posting to make a clear recognition of those that gave their heart and time to help create this wonderful event…to them, thank you. To our attendees of the event, thank you. Together, we all gain by working together to reduce energy use and costs.

This event and efficiency program would not exist if it were not for the vision and leadership of Dale Sartor and Bill Tschudi of Lawrence Berkeley National Laboratory since the inaugural 2008 Summit. Under their direction, LBNL has led the development of this program with funding from the California Energy Commission and its PIER program, which funded some of the project management and some of these technology demonstrations.

As my favorite business school professor always, said “the most prepared team always wins.” The Silicon Valley Leadership Group, particularly Bob Hines and staff Anne Smart, Samantha James, Colin Buckner and others ensured that the preparations were in place for our event today. Thank you.

Thank you Brocade, our host for the 2010 Data Center Efficiency Summit, especially John Noh, Rebekah Johnson and Brocade’s Events planning team.

Thank you to my co-chairs, I could not have pulled this event off without any of you: Tim Crawford, Ray Pfeifer, Kelly Aaron and Brian Brogan. And also our volunteer Summit Planning Team of Subodh Bapat, Henry Wong, Deborah Grove, Mukesh Khattar, Patricia Nealon, Ralph Renne, Joyce Dickerson, Zen Kishimoto and Chris Noland.

And most importantly, kudos to all of our speakers, moderators and panel members for sharing your knowledge, vision and ideas; we all have gained from your selfish tutelage. Let me know what you’d like to see for a future summit. Now go save some energy and let’s all have a drink together to celebrate today’s success!”


2010 Data Center Efficiency Summit a Great Success

Monday, October 25th, 2010

Well folks, the 2010 SVLG Data Center Efficiency Summit on October 14th was a raging success. As the Co-Chair of this event, I am thrilled that we had about 530 people at an excellent venue that was set to accommodate 450 seats. Brocade did an excellent job hosting the event. I am proud that we had 18 excellent end-user case studies and several other detailed discussions about pertinent industry topics by many top industry thinkers. Here are some photos of the event. To all that attended, thank you for coming. You can directly tell me your feedback of the event. Case studies will be posted on the website and I’ll post some on my website as well. You can also view posts of the event on my In the News page. Here is a link to a funny blog from David O’Hara about a session that I moderated.

Switzerland for Yahoo! Data Center

Monday, October 11th, 2010

Yahoo announced that they are building a data center in Avenches, Switzerland. This is a project I worked on over two years ago while I was with Yahoo! I don’t talk about material non-public information from my ex-employers or clients (even if unprotected by an NDA), so now that it is public, I’ll share some more about this site selection and strategy. As Yahoo! has announced during their 2010 investor presentations, in 2007 (when I started as an employee at Y!), over 90% of data center capacity was leased. Due to the strategy I and others put into place and is being implemented through these data center builds, Yahoo! estimates that about 90% of data center capacity in 2013 will be owned, providing a 35% total lower cost by 2014. It is these lower costs that drive companies to build their own data centers, and certainly Yahoo! has the economies of scale to make this effective. (Coincidentally, the lower total costs also makes hardware have a larger cost percentage, proving James Hamilton’s (Amazon) point that server costs are the biggest cost item of data centers.)

Some may ask why put a data center in Switzerland? Well, there are many good reasons to choose Switzerland (CH) for a data center in Europe. It does have higher energy prices than Russia, Ukraine and Finland, as well as the ex-Baltic States–Lithuania (half of my heritage), Latvia and Estonia–yet low for the rest of Europe. With over 1,000 small utilities, energy prices can vary much throughout the country. We can rule out Russia and Ukraine even though they have low energy prices for challenges to conduct business especially building complex data center projects. The ex-Baltic states can be excellent, but there are limitations to the infrastructure in some areas, so not as easy to complete projects. While I was with Google we looked at locations in this area.

CH has excellent fiber all around the country. It’s in the middle of Europe, so latency is superb to all of Europe. Work forces are superb with people speaking multiple languages (English, French, German and a local dialect are national languages) and very educated folks. Construction quality is probably the best in the world, with construction and engineering firms from all over Europe providing skills. Geneva is a one-hour flight from much of Europe including England, France and Germany. Train service is superb, with quality trains, regular schedules, low fares, precisely on time departures and arrivals, making driving seem mundane and so archaic.

A lovely “bonjour” welcomes you everywhere you go. And the Alps seem to surround CH with gorgeous green countryside and quaint small towns. Avenches, where we chose to put this Yahoo! data center, is one of those quaint towns, with a Roman amphitheater and other Roman features and history, a small lake and lovely green fields. Plus, traveling in thru Geneva and then onto Lausanne it beautiful, across the lake from the French town of Evian of the water fame, with reflections of snow-covered Alps glistening off the lake. Can you tell I thoroughly enjoyed working on this project? The food is tops in Europe, the air is clean, the water is clean, the cities are clean, and the power is clean. Avenches even has it’s own district heating loop from a small bio-mass plant. With most of CH’s electricity generated from nuclear and hydro, CH has one of the lowest carbon impacts per kWh anywhere in the world, and this is important to Yahoo! as they strive to be carbon neutral, a very difficult achievement in a growing company with many data centers around the globe.

So CH is clean, beautiful, easy to do business, central to Europe, safe, close to Chamonix and Zermatt for skiing side trips on weekends and it has low taxes, some of the lowest in all of Europe. Making CH a great place to put corporate headquarters as well. The latest Global Competitiveness Report places CH tops in their overall ranking:

“Switzerland tops the overall ranking in The Global Competitiveness Report 2010-2011 released by the World Economic Forum. The United States falls two places to fourth position, overtaken by Sweden (2nd) and Singapore (3rd). The Nordic countries continue to be well positioned in the ranking, with Sweden, Finland (7th) and Denmark (9th) among the top 10, and with Norway at 14th. Sweden overtakes the US and Singapore this year to be placed 2nd overall. The United Kingdom, after falling in the rankings over recent years, moves back up by one place to 12th position.”

I’ve worked on data center projects in all top-10 countries and most countries in Europe. Switzerland is excellent but so is Sweden as they have been working hard to lower taxes, lower energy rates, and energy is VERY clean with low ambient temperatures and lots of low temperature water for direct cooling to reduce energy use. Sweden actually has fairly low effective tax rates, much lower than one would think of a Nordic country. Switzerland is great and so is Sweden—think fiber access into the ex-Soviet block.I predict that we’ll see more new data center developments in the Nordic countries, ex-Baltic states and some in Switzerland, as there’ll be a gradual transition from the key European cities for data centers (London, Amsterdam, Frankfurt & Dublin) to mostly Northern and Eastern Europe for energy costs, lower energy use and lower tax cost benefits. Also to reach that large Russian Internet market.

I’d love to share more about my experiences and knowledge of low-cost, excellent places to put data centers in Europe. Call to schedule a time to discuss your European expansion. In the meantime, enjoy that croissant. Au Revoir!

Equinix has a Bad Market Day

Friday, October 8th, 2010

Equinix had a bad day Tues, with stock price dropping by about a 25% in one day and not recovering since. This dramatic price drop also affected the stock value of many prominent public data center companies or data center centric companies. The Equinix price drop was created from a revenue warning from Equinix and a lawsuit filing questioning churn and revenue forecasts. The market is clearly jittery about any uncertainty in the revenue filings and forward expectations of margin and revenue for what has been a very robust company and market segment. These questions have hurt all companies in this segment. My analysis, and this is purely my analysis and opinion as an industry person, not a recommendation or investment professional report.

Equinix has always reported churn to be very low and that it’s customers have a certain stickiness due to its collection of network providers at each site and the group of core customers attracted to be in Equinix for peering with each other. And while I have been saying for years that fiber is ubiquitous, adding insignificant dollars to large data center projects to bring fiber to even remote sites, as a past large customer of Equinix, I can attest to the value of these peering relationships and network options. However, while there is great value in these network connections, they can be accomplished with little equipment, a rack or two. Meanwhile, the rest of the customer’s servers are located in a lower cost wholesale or internal data center (I’m referring to large users). And as networking technologies and equipment improve and shrinks in size to accomplish the same number of connections, “network” customers of Equinix need fewer racks for these networking connections. A past 2-4 rack requirement is now 1-2 racks, so churn is not loss of customer but a decrease in the number of racks. Meanwhile, that customer may add more cross-connects to peer with more Equinix customers in the same data center, further entrenching the “network” community that is there to peer with each other. And each of these cross connects is fairly low monthly recurring revenue (MRR) yet pure margin, so Equinix maintains high margin even if total MRR declines slightly over time. Overall, a good business model that retains high margins even if revenue falls slightly.

The challenge for Equinix is to grow MRR as customer equipment shrinks, since Equinix sells mostly space in a power driven world. Also wholesale providers are selling to customers with 20+ racks instead of what was the minimum buy in at about 100 racks. So the wholesale and other data center providers are closing in on the smaller rack customers that are core to Equinix while Equinix has not moved up in size to competitively serve the larger rack customers, those with over 25+ racks, as the wholesale providers serve those customers at significantly lower costs. As large rack customers of Equinix migrate to wholesale providers, they keep up to a few racks to maintain networking connections with high margin but Equinix looses market share and larger revenue customers, essentially driving Equinix to their core business and core high margin customers with a few racks and lots of cross connects, exactly what they want purging the large customers with dozens of racks at a lower MRR per rack. So while rack churn sets in customer stickiness remains. Summary point: Network is ubiquitous; most don’t need 20+ carrier connections; those that do keep a networking rack or two for peering and networking connectivity and maybe caching servers; rest of racks move to lower cost centers

Separately, cloud providers continue to move in to Equinix, an excellent growing customer segment for Equinix. But I predict that as these providers get large in the number of racks, where it becomes lower cost to relocate to a wholesale or other lower cost provider, they will. I used to work for Exodus Communications who started the co-location market and I also ran a data center co-location company several years ago, and as a past large customer of Equinix, DRT, DFT and many others, and also ran data center operations for one very large Internet play and once very large software provider, so I am familiar with the challenges of moving racks from one location to another. This challenge adds to the stickiness of customers and limits churn. But as customers grow and add new servers, customers can add new servers in new locations via cloud and with the help of virtualization and quickly have them virtually replace servers else where, making transitioning from one data center to another a much easier task, whereby primary servers can change from one data center location to another, and then be physically taken offline, making data center transitions much easier than before. This is especially true as more and more folks do not rely on the one data center basket for all servers, and instead, of multiple, geographically diverse data centers that load balance and fail over to each other in real time.

A couple of final points. I like Equinix; I believe they have an excellent product and a very high margin business. While being very focused to retain margins they have missed what I believe are revenue growth opportunities in booming Asia and with larger rack customers albeit at lower margin. Also, the intense focus on networking hubs has forced them to have data centers in high power cost markets, which directly and significantly affects COGS while corporate data centers have been moving out of metro areas to rural areas for cost benefits. This brings better fiber to those areas and “known-ness” to those areas as we see with data center bunching, for better of for worse in places like Central Washington, Utah, Reno, North Carolina, etc. Equinix is limited by their focus on network and high margin to metro areas, again further limited revenue growth and having a higher impact from high power costs and market share loss to those moving to lower power cost centers. This will help them retain margins while likely loosing market share, so I see Equinix’s value in high margin with less revenue growth—a result that should be valued but we’ll see what transpires over time, as many investors have often been more focused on revenue growth. Let’s hope all data center stocks are not devalued and that revenue forecasts are accurate as possible to maintain the proper value of these data center companies.

You can read more via these fine articles here and here:

Also, be sure to register for the SVLG Data Center Efficiency Summit on October 14th in San Jose, as registrations are expected to be sold out soon.