Friday, October 3rd, 2014

Time	Event
12:00p	No Stone Unturned There are many ways to reduce data center power consumption, and Billie Haggard’s team at CoreSite Realty Corp. leaves nothing off the table when it comes to optimizing their 16 U.S. facilities for energy efficiency. CoreSite and its biggest competitor Digital Realty Trust are the only data center providers that have joined the White House’s Better Buildings Challenge. Each participant in the challenge commits to reducing (verifiably) energy consumption of a set of buildings in their portfolio by 20 percent between 2011 and 2020. Earlier this week the Department of Energy, which oversees the program, announced that 19 organizations (including CoreSite and Digital Realty) have joined it specifically to reduce energy consumption of their data centers, as opposed to other types of buildings. The only other private-sector entities on the list were Schneider Electric, eBay, Staples and Home Depot. The others were government organizations. A data center efficiency think tank CoreSite has had its own portfolio-wide energy efficiency program since 2010, Haggard, senior vice president of data centers at the company, said. The program is strategic in nature, a part of CoreSite’s business plan. Haggard’s primary reason for joining the White House program was the opportunity to work as part of a think tank, together with data center heads of government agencies and other companies. There are benefits in sharing knowledge, resources and ideas about data center power efficiency, he said. After working in the data center industry for 14 years, however, he is probably more a mentor than a student in the group. Before CoreSite, Haggard worked for Lee Technologies, a provider of technical services to data center operators acquired by Schneider in 2011, and Switch and Data, a colocation provider Equinix gobbled up in 2009. No stone will be left unturned Since the Green Buildings Challenge allows participants to use energy consumption data from 2011 on retroactively, Haggard’s team at CoreSite is planning to include improvements they have made over the past three years in the energy reduction calculations they present to the DoE. CoreSite chose 10 of its 16 facilities to participate in the challenge. Past and future improvements can be everything from replacing inefficient UPS systems and cooling equipment to fine-tuning chillers or implementing data center infrastructure management software. “We may find that operating three chillers at 33 percent [of total capacity] is more efficient than operating a chiller at 100 percent,” Haggard said. Another trick is expanding use of waterside and airside economization (known as free cooling) hours. The latest ASHRAE guidelines for data center IT equipment allow for warmer ambient temperature on the raised floor than before, so data centers can use free outside-air cooling in more locations and over longer periods of time. CoreSite has also been implementing a DCIM solution across its facilities. Echoing a common sentiment, Haggard said there wasn’t a DCIM product on the market that did everything his team wanted it to do, so they’re using a combination of off-the-shelf DCIM tools and proprietary home-baked software. “There’s not a single DCIM package out there that meets all of our needs,” he said. Replacing old-school humidifiers that work by boiling water to turn it into steam with modern ultrasonic ones is another big source of savings. The old technology in a computer room air handler could require as much as a 50 Amp breaker, while the ultrasonic alternative (which vaporizes water by vibrating it at ultrasonic frequencies) needs 20 to 30 Amps, Haggard said. How much of it will be a ‘paper exercise?’ Together, the 19 organizations that recently joined the Better Buildings Challenge represent more than 90 megawatts of data center power, according to the DoE. While there are only two data center operators participating, they have some of the largest real estate portfolios in the country. Digital Realty has not yet identified the block of facilities it will use for the challenge, but if it ends up being of similar size to CoreSite’s block, a 20-percent energy use reduction across that many big data centers can make a big difference. However, neither company has said how much of that energy-use reduction will have come from improvements they have made already, regardless of the White House challenge. David Schirmacher, senior vice president of data center operations at Digital Realty told us yesterday he was not planning to do any “paper exercises” to get to the 20 percent. How much of the effort on every participant’s part will not be a paper exercise is what will determine the actual impact of their participation. (Comment on this)
3:30p	Scaling Your Network for Big Data Michael Bushong is the vice president of marketing at Plexxi. You might be surprised to hear that for most companies, Big Data is not really all that big. While some Big Data environments can be massive, typically deployments are 150 nodes or less. The technology and applications are still fairly nascent, so outside of a relatively small number of players for whom Big Data underpins their business, most are just dipping their toe in the proverbial Big Data waters. But just because the initial foray into Big Data is small, doesn’t mean that architects building out infrastructure for a Big Data era can ignore scale altogether. Scale, but not the way you would expect In the networking world, the word ‘big’ evokes visions of massively scaled out applications spanning thousands of nodes. Based on the amount of data replication happening across these nodes, the interconnect that pulls together all of these resources will surely be enormous. Or will it? While some web-scale properties are likely to develop and deploy Facebook-like applications, the vast majority of companies are unlikely to build their businesses around a small number of applications with this kind of scale. Instead, as initial experiments in Big Data yield fruit, companies will roll out additional small-to-medium sized applications that will add incremental value to the company. Perhaps an application that performs well for one line of business will be picked up by another line of business. Maybe a recommendation engine that yields additional retail revenue will be augmented with an application that looks at buyer history. One by one, these applications add up. If a single small application has 150 nodes, even six to eight applications start climbing into the 1,000-1,200 node range very quickly. If you consider that Big Data is really just a specific type of clustered application (along with clustered compute, clustered storage and any distributed application), the number of potential applications to consider grows even higher. Competing requirements Scaling for multiple applications is an entirely different beast from supporting a single massive application. With a single application, the requirements are uniform, allowing architects to design the networks with a specific set of objectives in mind. But, different applications have different requirements. Some applications are more bandwidth heavy. Others only run at certain times. Some are particularly sensitive to jitter and loss, while others have strict compliance (HIPAA and PCI) requirements. The challenge is that when you first design the network, you don’t know which of these requirements will ultimately hit your network. This problem is only exacerbated when you look at how many of these smaller projects are initiated with Big Data in mind. For instance, a line of business spearheads the first move toward Big Data. They fund a couple racks of equipment, and connect them using some fairly nondescript networking equipment. It’s more of an experiment than a core necessity for the business, so the work is done on the side. This protects the rest of the network from the potential issues related to retransmission storms. The approach of isolating new deployments makes sense, but it makes it easier to plan only for the short term. As each deployment is optimized for a specific application, architectural assumptions become narrower. When these short-term experiments evolve to business-critical features, converging multiple applications that exist in specialized or niche environments can be prohibitively painful. The only remediation is to continue with dueling infrastructures for each class of application, or to take on the burdensome task of converging on a common infrastructure. Planning for convergence Ultimately, the key to transitioning from the role of Big Data experimenter to that of Big Data practitioner is to plan from the outset for the eventual convergence of multiple applications with competing requirements onto a common, cost-effective infrastructure. The networking decisions at the outset need to account for a set of rich applications with different needs. This leads to a set of questions that architects should consider: If applications require different treatment, how are those requirements communicated to the network? Is the network capable of altering behavior to optimize for bandwidth requirements? Latency? Jitter? Loss? Can application traffic be isolated for compliance and auditing reasons? If growth over time leads to physical sprawl, how are issues with data locality handled? Do applications suffer if resources are 100 meters apart? 100 kilometers? When a cluster grows from 100 to 1000 nodes, what are the cabling implications? Do you have to change out cabling plants? What are the operational implications of implementing this diverse network? Network scaling for Big Data is more than just planning for the eventual number of nodes in a Hadoop cluster. Careful consideration for how to gracefully grow from one-off, bite-size deployments to a rich set of diverse applications is of paramount importance to avoid the types of unplanned architectural transformation projects that can wreak havoc on budgets and cripple IT departments for years. Plotting a deliberate architecture course from the outset is the surest way to guarantee success. Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library. (Comment on this)
4:30p	Friday Funny Caption Contest: Blue Screen It’s hard to believe, data center friends, but October is upon us again. Let’s get in the Halloween spirit with this week’s Friday Funny! Diane Alber, the Arizona artist who created Kip and Gary, has a new cartoon for Data Center Knowledge’s cartoon caption contest. We challenge you to submit a humorous and clever caption that fits the comedic situation. Please add your entry in the comments below. Then, next week, our readers will vote for the best submission. Here’s what Diane had to say about this week’s cartoon, “I saw a costume like this just the other day and thought it had Kip’s name written all over it!” Congratulations to the last cartoon winner, Rodney, who won with, “The ghost servers are going to the cloud now . . .” For more cartoons on DCK, see our Humor Channel. For more of Diane’s work, visit Kip and Gary’s website. (Comment on this)
5:03p	Microsoft Unusually Candid With Next Windows Server Preview With the consumer edition of Windows leapfrogging version 9 for version 10, Microsoft surprised many by announcing Technical Preview builds of yet-to-be-named editions of Windows Server and System Center. Released on October 1, these very early previews are being provided, uncharacteristically, to a broad audience for testing. The last Server release was code-named Windows Server 8, so perhaps the next release will follow suit and become Windows Server 10. Although it has not given much detail on this, Microsoft said it will change how platform products will be released. The only hint given was that the company is trying to deliver the best of both worlds – for those who favor stability and predictability and for others who want access to the latest and greatest technologies as fast as possible. Microsoft knows a thing or two about running global-scale clouds and data centers, and its Server and Cloud Platform team says they are taking two core components of that cloud platform, Windows Server and System Center, and bringing them to the new releases. Add Azure into that mix and you have the foundation of the Microsoft Cloud OS vision. One of the interesting features listed on the future Windows Server is support for rolling upgrades of Hyper-V clusters without downtime for workloads. New components for its SDN stack feature a network controller role to manage virtual and physical networks. Microsoft says new synchronous storage replication plus storage Quality of Service will help deliver minimum and maximum IOPS in environments with workloads with diverse storage requirements. Windows PowerShell advances to version 5.0, with new security features, support for developing with classes and new features in Windows PowerShell Desired State Configuration (DSC). Some of the more Azure-influenced changes include replacing the App Controller with Windows Azure Pack, Server App-V support replaced with template-based migrations and Cloud Services Process Pack (CSPP) replaced with Windows Azure Pack functionality. The company is planning to discontinue support for the old Windows Server 2003 in July 2015. Microsoft said last month that it estimated that nearly 40 percent of the entire Windows Server install base was still that 2003 version. (Comment on this)
5:30p	Data Foundry “Tops Off” Houston Data Center Data Foundry held a “topping off” party last week and provided a look into progress at its upcoming Houston data center. Construction of the 350,000 square foot facility is on track: the roof is on; the generators are in. The company will start tours very soon. In builder jargon, “topping off” a building means placing the last beam on top of a structure during construction. The data center will be the company’s largest greenfield project to date. Construction crews broke ground at the site in April. Houston 2 resides on an 18-acre tract of land the company purchased in 2013. The new data center is modeled after Texas 1 built in Austin in 2011, which has been successful for the company. “Austin was a very small data center market. We created a much larger market with large-scale product,” said Edward Henigin, CTO, Data Foundry. Houston is a much bigger market than Austin and the company reports solid interest. Other providers in Houston include CyrusOne and SoftLayer (IBM), as well as new wholesale entrant Skybox (which has partnered with T5 for data center management ). Houston is home to a total of about 30 data centers. But market watchers say supply is tight, and the large footprints of recent deals in the energy sector suggest continuing demand. Before: the progress last May, a month after the ground breaking (source: Data Foundry) The walls go up on Houston 2 (source: Data Foundry) Data Foundry is part of the larger trend of blurring lines between wholesale and retail colocation. Houston 2 will serve requirements ranging from single-cabinet to multi-megawatt deployments. “The retail customer, you move in, everything works,” Henigin said. “It comes out on time every time, like Starbucks. We have that level of finish. We can sell a metered deal with the wholesale market or we can sell individual racks.” Henigin said the provider’s experience and expertise were real assets to customers. “It is a balance between providing a high level of security while at the same time providing those creature comforts in the data center,” he said. “People are more productive when they have lower stress.” The underground chillers (source: Data Foundry) (Comment on this)
5:51p	IBM Makes OpenPOWER Play Against Intel IBM is doubling down on its open server innovation strategy through the OpenPOWER Foundation. The company announced a new system incorporating OpenPOWER to better address Big Data challenges. The company is aiming squarely at Intel‘s stranglehold on the server processor market, claiming it has achieved better price performance than comparable Intel-based servers. IBM, Google, NVIDIA, Mellanox and Tyan founded the OpenPOWER Foundation in 2013. It has attracted 59 members to date, all reportedly working together to leverage POWER’s newly opened server architecture. The momentum is strong. The org has grown fivefold in nine months. IBM announced a $3 billion investment in research and development of processor technologies in July. The investment indicated IBM had no intention of abandoning its chip business altogether, despite the sale of x86 server business to Lenovo. There were news reports that cited anonymous sources saying Big Blue was considering a sell-off of the processor business. But IBM has a long way to go if it wants to suck meaningful market share from Intel, which powers majority of the world’s servers. Commenting on July’s announcement, an Intel spokesperson told us Intel was spending $10 billion a year on processor R&D. “Open” but pay-to-play IBM is investing in technology development rather than manufacturing. It licenses the POWER architecture to others, but is itself a chip technology licensee. IBM licensed processor architecture from UK’s ARM Holdings last year, saying it would use it to build custom chips for clients. This is arguably a battle of business models — at least IBM is positioning it that way — closed and proprietary versus open. But IBM’s approach with POWER isn’t open in the same sense open source software is open. It has made POWER server technology open through the foundation, but it sells licenses for the IP. It means companies are free to innovate on the tech, but they still have to pay the owner. The strategy behind selling x86 business to Lenovo is now clearer. IBM is putting its weight behind OpenPOWER technology instead of x86 server manufacturing and waging war against leader Intel with the “open” angle. “Our open innovation business model and approach to OpenPOWER will disrupt technology providers that offer closed, proprietary solutions produced within the walls of one company,” said Doug Balog, general manager of Power Systems at the IBM Systems and Technology Group. The wares The new IBM Power S824L servers are built on IBM’s POWER8 processor, which the vendor says is optimized for Big Data workloads. Built on the OpenPOWER stack, the systems provide the ability to run data-intensive tasks on POWER8 processors while offloading other compute-intensive workloads to GPU accelerators. GPU accelerators are capable of running millions of data computations in parallel and are designed to speed up compute-intensive apps. IBM is working to optimize Power versions of widely used GPU-accelerated applications for bioinformatics, defense, finance, molecular dynamics, weather modeling – including SOAP3, NAMD, GROMACS, FFTW library and Quantum Espresso. Future versions of IBM Power Systems will feature NVIDIA NVLink technology, eliminating the need to transfer data between the CPU and GPUs over the PCI Express interface. This will enable NVIDIA GPUs to access IBM POWER CPU memory at its full bandwidth, improving performance for numerous enterprise applications. These systems are expected to be available beginning in 2016. A post-x86-deal pivot The IBM x86 server deal with Lenovo was a long time in the making, IBM first announcing intentions to sell in January. The deal positioned IBM to make its big OpenPOWER push. Lenovo was also a victor, becoming the third-biggest x86 server provider worldwide and the leader on the key turf that is China. “This divestiture allows IBM to focus on system and software innovations that bring new kinds of value to strategic areas of our business, such as cognitive computing, Big Data and cloud,” said Steve Mills, senior vice president and group executive, IBM Software and Systems. (Comment on this)
6:30p	Cloudera Buys DataPad’s Python Chops In a move to strengthen its enterprise data hub Cloudera announced it has acquired the technology assets of data visualization company DataPad , a startup with deep expertise in using Python for data analytics. San Francisco startup DataPad launched last year with the vision from co-founders Chang She and Wes McKinney to build better data tools for an integrated, accessible data discovery environment. McKinnery created the open source Python Pandas project, a software library for Python-based data manipulation and analysis. DataPad launched with a $1.7 million investment from Accel Partners, Google Ventures, Andreesen Horowitz and Ludlow Ventures. DataPad offers automated analytics that connect to a variety of data sources and present visualizations with options for collaboration and optimizations for touch-screen devices. Big Data analytics and data visualization for Hadoop is an active market right now, and the Cloudera acquisition seems to have been primarily about bringing the data engineering talent from DataPad into the mix at Cloudera. DataPad specializes in data analysis using the Python programming language. With Python-based tools added to Cloudera’s Big Data management and analytics platform the company looks to reach more developers and data scientists and further expand its contributions to open source projects. Apache Spark was added to Cloudera’s enterprise data hub earlier this year, promoting the notion of fast data that is instantly actionable. Cloudera went on to write a Python client for the open source Hadoop query engine Impala. “We are thrilled to have the DataPad team join Cloudera and look forward to their contributions to the Cloudera roadmap,” said Peter Cooper-Ellis, vice president of engineering at Cloudera. “We’ve long been supporters of the DataPad team and have been impressed with their engineering work. Together, we possess some of the best talent in the data engineering sphere. The deep Python expertise that DataPad brings to Cloudera will further accelerate our data engineering capability.” (Comment on this)
7:00p	Online Tech Steps Outside of Michigan Online Tech’s Indianapolis data center is open for business after the company completed several upgrades. The 3-megawatt facility’s acquisition and upgrades amounted to a $10 million investment. The 44,000-square-foot data center is the company’s fifth data center and its first outside of Michigan, where it operates four facilities, totaling 100,000 square feet of space. Online Tech is targeting compliance-minded businesses, a market it says is underserved in Indianapolis. The company announced it had acquired building in May. It was previously used as a data center but required significant work, according to the provider. “One of the first upgrades we made was replace the power infrastructure circuits,” Yan Ness, an Online Tech co-CEO, said. “We also installed new PDUs that meet our needs. The company also installed a new hardware infrastructure to support cloud services. The setup includes EMC storage and Cisco UCS servers and network equipment. The infrastructure enables Online Tech to move and share workloads between its Indianapolis and Michigan locations. “That will allow us to do some cool things for customers,” Ness said. The company performed physical remodeling as well, focusing efforts on things that met customers’ non-technical needs. “We wanted the facility to be more than just a data center for customers. We wanted it to be their IT shop and their home away from home, so we built in meeting rooms, office space, a kitchen and lots of other amenities that make the space really livable as a day-to-day office space and meeting space for them,” Ness said. (Comment on this)
7:31p	Cloud Reboot Causes Cold Sweat at Netflix Another tale has emerged from the great server reboot of 2014 to apply a Xen security patch that affected major cloud providers, including Amazon Web Services and Rackspace. Netflix, an AWS customer, lost 218 database servers during the reboot but managed to stay online. Last week, a known issue that effects Xen environments forced AWS and Rackspace, in addition to other Xen users, to reboot portions of their clouds to apply a Xen security patch. It was done on short notice to customers. Luckily, the reboot went smoothly for both. Now a major customer story has emerged. Netflix was concerned about its Cassandra database. The company has over 2,700 production Cassandra nodes, of which 218 were rebooted. A total of 22 of those were on hardware that did not reboot successfully. However, Netflix’s automation detected the failed node and replaced them all with minimal human intervention. Chaos Monkey, the company’s homegrown tool for testing resiliency, had already wreaked potential havoc on the system in the past. It didn’t mean, however, that the company was not concerned. “When we got the news about the emergency EC2 reboots, our jaws dropped,” said Christos Kalantzis, Netflix manager for cloud database engineering. “When we got the list of how many Cassandra nodes would be affected, I felt ill. Then I remembered all the Chaos Monkey exercises we’ve gone through. My reaction was, “Bring it on!” Netflix has a massive AWS infrastructure. The streaming movie provider has perfected the art of resiliency through its “Simian Army” resiliency tools. Chaos Monkey is a resiliency tool that randomly disables virtual machine instances that are in production on the Amazon cloud. The goal is to engineer applications so they can tolerate random instance failures. Chaos Gorilla disables all Netflix infrastructure in an AWS Availability Zone. (Comment on this)

<< Previous Day

2014/10/03 [Calendar]

Next Day >>