Defining the Future of IT by mirit35


									Sharing beSt practiceS with the information technology community winter 2009


Defining the Future of IT
Technology innovation for the next 40 years

also inside Crossing the Chasm Between Humans and Machines
page 8

intel enterprise Roadmap
page 58

supercomputing Gets Personal
page 24

article reprintj

The Future Will be simulated Small business units need high performance computing. to meet that demand, cray and intel have partnered on a transformative solution. Join the Intel Premier IT Professional program




Winter 2009
B:11.25 in S:9.75 in
cover illustration by john macneill

Crossing the Chasm between Humans and Machines Intel’s CTO Justin Rattner shows developers leading-edge evidence of the narrowing humanmachine interface that’s driving the advance to what futurists call “The Singularity.” Making the Most of IT The Innovation Value Institute’s grand vision: to transform the way enterprises get value from IT through a proven and validated best practice model for IT.

36 Virtualization test Intel Xeon processor 7400 series—code-named Dunnington—holds great promise for improved performance in virtualized systems. 42 Out with the New, In With The Newest Intel IT has shown it makes more sense to refresh design center servers every four years rather than wait until they’ve outlived their usefulness. 46 The Fast and The Virtual Refreshing client PCs helps fulfill the promise of client virtualization.

20 Intel’s New Approach to Strategic Planning IT leaders within Intel develop a long-range vision for improving IT efficiency. 24 The Future Will be Simulated Small business units need high performance computing. To meet that demand, Cray and Intel have partnered on a transformative solution. 30 Mad Dogs and Oil Men In BP’s High Performance Computing Center, each new generation of microprocessor helps seismic researchers strip away the mystery shrouding fields of oil.

58 Intel Enterprise Roadmap Intel’s Product Roadmap, in easy-to-read graphic format. 62 Mashups for the Masses Intel’s Mash Maker, a browser extension, allows users to pull information from Web sites to create “mashups,” a custom browsing experience. 66 Intelpedia Grows Up Intel’s wiki has, in just under two years, grown to nearly 25,000 pages and houses a huge variety of information about the company and its products.

premier it | winter 2009 |

T:10.5 in


16 The Value Influence For Ramon Baez, Kimberly-Clark’s CIO, partnering and leadership add up to Big Impact.

54 It Can be Easy Being Green Facing Expansion, Oregon data center leverages Intel and a host of vendors and technologists for energy-efficiency.



Experiments lead to scientific discovery. But experiments can be extremely expensive, time-consuming, and prone to dead ends. That’s where computer-based simulation comes in, unearthing the most promising areas for development without having to rely on real-world trial and error.
Executive Summary
Cray introduces a new category in highperformance computing: the personal and workgroup supercomputer.

The power of simulation also holds true for the commercial enterprise. Simulation techniques such as virtual prototyping and large-scale data modeling can shorten the time it takes for a company to get a successful product to the marketplace, while helping to ensure high quality and further new product development. The competitive benefit of simulation has driven incredible demand for high-performance computing (HPC). “HPC has become a critical technology for any high-tech company,” says Earl Joseph, program vice president of high-performance computing at the research firm IDC. “In the last five years, HPC has grown faster than any IT sector, growing roughly on the order of 20 percent per year.”

premier it | winter 2009 |

The Future Will Be

To design and manufacture the products of tomorrow, small business units need highperformance computing today. Cray and Intel have partnered on a transformative solution.



premier it | winter 2009 |


premier it | winter 2009 |

Inside the Intel-Cray Partnership
Through a recently announced R&D partnership with Intel, Cray has ensured the future convergence between the respective technology roadmaps of the two industry leaders. In the short term, CX1 customers will be able to upgrade individual blades using new generations of Intel processors as soon as they become available. Looking ahead, future generations of Cray personal and workgroup supercomputers will take advantage of improved Intel designs—and vice versa. Future innovations in the Cray CX1 pipeline include symmetric multiprocessing (SMP) shared memory software for memory-intensive applications; and a quadsocket compute node based on the Intel Xeon processor 7300 series, with more memory and enhanced capabilities for field-programmable gate arrays (FPGA), which allow software developers to reprogram the chip on-the-fly to improve application-specific

performance. “The CX1 is a class of machines that will take advantage of Intel Xeon processor technologies,” says Richard Dracott, general manager of High-Performance Computing at Intel. “Beyond today’s leadership performance from the Intel Xeon processor 5400 series, you will see additional improvements for HPC workloads in 2009.”



premier it | winter 2009 |

But so far, adoption has been limited by the need for specialized talent and custom software. While the top industrial performers use the most advanced supercomputers and HPC clusters to perform the most comprehensive simulations, it has been difficult to scale down HPC to smaller firms. Setting up large HPC clusters is a complex and daunting task, requiring knowledge of how to specify, configure, and maintain a complex infrastructure across multiple vendors. As a result, penetration of HPC has been largely limited to the leading U.S. companies and best-in-class international firms with the scale and expertise to manage complex HPC deployments. Although HPC has become a competitive necessity for the largest firms, it remains a relative rarity among smaller companies in the United States. Some small companies overseas, particularly in East Asia, have nevertheless succeeded in deploying HPC to drive product improvements throughout the value chain. “Through simulations, subcontractors are telling their customers where they need to change their designs,” explains Joseph. “They can show where the stress points are, where they should make the metal a little thicker, where the gears need some extra case hardening and so on, to provide a better product.” For instance, a team of engineers using HPC to design windshield wipers can model the effectiveness of the blades under various weather scenarios, help-

the team ends up with a shared server under central IT control, gaining computing power at the expense of the flexibility and customization that scientists, engineers, and designers have long enjoyed on their personal workstations. After all, it’s not as if a small group of engineers or designers could simply go out and buy their own Cray* supercomputer— until now.
Introducing the Cray CX1 With the Cray CX1, supercomputing pioneer Cray has introduced a new category in HPC: the personal and workgroup The Cray CX1 is part of a new category of supercomputer. “People supercomputers. want more power than they’re getting from even the most powerful desktop ing them to improve the design workstations that you can buy beyond the specifications of the right now,” says Jeff Cachat, CX1 automaker. This type of computeprogram manager at Cray. “The intensive work requires only a CX1 bridges the gap between the fraction of the power of a fullworkstation and the traditional fledged supercomputer, yet it’s rack-mount cluster.” beyond the capabilities of a workIn the form factor of a deskstation. side system, the CX1 consists of Teams caught in the middle a single preconfigured cluster between workstations and superwith room for up to eight separate computers have been stuck with blades, which buyers can mix and an unappealing choice—either match to meet their specific busicope with the bottlenecks resultness needs: ing from their underpowered p Computational power: Installmachines, or overinvest in expening eight dual-socket compute sive HPC clusters not suited to blades puts 16 Intel® Xeon® the smaller workgroup. Setting up processors (quad-core) within a a server room for an HPC cluster single cluster. Scientists, engiinvolves capital investment in neers, and analysts can use this physical plant and equipment, power to tackle the most comthe need to hire specialist techniputationally intense problems, cians, and difficult constraints whether in financial risk manover power usage and space utiliagement, genetic engineering, zation having long-term implicaor climate change simulation. tions. At the end of the journey,

p Memory throughput: Singlesocket compute blades have smaller memory capacities than dual-socket blades (16 GB RAM for single-socket versus 64 GB RAM for dual-socket), but they’re significantly faster when it comes to accessing that memory (up to 1,333 MHz for single-socket versus 800 MHz for dual-socket). For applications where speed to memory access is the biggest bottleneck, such as with trading applications in the financial markets, single-socket compute blades may be the best choice. p Visualization: Combine a dualsocket compute blade with an NVIDIA Quadro* FX graphics card, and you have a visualization blade that can quickly render complex graphics for simulating airflow over a wing, generating 3-D product designs, or designing Hollywood movies. p Storage: Another useful combination joins a dual-socket compute blade with either four or eight high-capacity hard disk drives. This is ideal for database-driven applications

that require local journaling and redundancy features, since the storage blade minimizes the physical distance between data

It’s not as if a small group of engineers or designers could simply go out and buy their own Cray supercomputer— until now.

stored on disk and data stored in memory. These flexible configuration options permit a wide range of deployment possibilities. Up to eight people can share the CX1, with each person getting a dedicated compute blade as a replacement for separate workstations. As many as four animators can share a CX1 equipped with four visualization blades, providing each with a dedicated resource for

generating content and final output. A team of analysts can share access to a data warehouse. Or, for particularly demanding projects, a single power user can harness the capabilities of the entire machine. The CX1 puts powerful cluster technology in the service of small, decentralized, nimble business units. Designed to work in an ordinary office, the CX1 plugs into a standard 120/240-volt power outlet and thus does not have to be sited in the confines of a climatecontrolled and energy-constrained data center. With active noise cancellation technology, it’s just as at home underneath the desk as a traditional workstation. Furthermore, the software migration from the workstation is simple. The Cray CX1 offers two choices for operating systems: Microsoft Windows* HPC 2008, for an easy upgrade for Windows workstation users; or Red Hat* Enterprise Linux for a Linux-based deployment. Each solution comes with cluster management tools, including job scheduling, resource management, and system tuning utilities. That’s it. No dedicated server

premier it | winter 2009 |

rooms, no high-priced administrators, and no complex assembly required—just choose your blades and pick an operating system. This gives workgroups a quick and easy path to HPC. “This is a new way to bring Cray technology and capabilities into a much broader marketplace,” says Peter Ungaro, CEO of Cray. “With the lowest price point for a Cray product that we’ve ever had, we’ll be able to reach more customers with Cray technology than ever before.” Because the CX1 reaches into vertical markets outside Cray’s traditional market of supercomputer buyers, the company plans to work with partners to develop workgroup-ready solutions. Through the Intel® Cluster Ready program, software vendors and system vendors can certify the performance of the total solution using Intel processors, with benchmarks for easy comparison. For example, a software company with a computational fluid dynamics (CFD) solution may work with Cray to develop a benchmark showing how fast the application runs on the CX1, and then work with a channel partner or system integrator to deliver preconfigured Intel Cluster Ready CX1 units with the CFD software loaded and ready to go— giving customers in aeronautics, manufacturing, and the oil and gas industries the option to purchase a certified turnkey solution. Rather than having to learn how to build a HPC cluster from scratch, engineers will be able to focus on their core activities of simulating the behavior of fluids and gases around complex surfaces.
HPC’s High-Powered Impact The impact of the CX1 promises to be nothing less than transformative. Given that HPC already

makes up one of the fastestgrowing segments in the computer industry, the ability to order a preconfigured, preinstalled cluster in a single box makes it feasible to spread HPC to new industries,

For particularly demanding projects, a single power user can harness the capabilities of the entire machine.

regions and stages across the value chain. “There’s truly an insatiable demand for performance,” says Richard Dracott, general manager of HPC at Intel. “In every field where you see HPC being used, there’s always a clear understanding of what they’d do if they just had 10 times more performance— and 100 times and 1,000 times more.” Now imagine extending the use of HPC to companies that haven’t even considered the benefits yet. “Hundreds of thousands of small to medium-sized companies that don’t use workstations or HPC have a big opportunity here to get their products to market faster,” says Dracott. “People are starting to discover the value and potential of HPC, and with system integrators selling the CX1 as a turnkey customized solution complete with software, it’ll be as easy as buying a workstation.” Regionally, the Americas have been the largest consumers of HPC, but that’s no basis for domestic complacency. “Other

geographies are starting to increase their investment in HPC at a faster rate,” says Dracott. “To become more competitive at a global level, American companies will have to catch up.” Indeed, companies with a global workforce will benefit from the ability to distribute HPC resources to match their labor footprint. Rather than having to relocate personnel to the nearest data center or rely on dodgy Internet connections, firms can deploy the CX1 to worldwide locations. “Several digital content creation companies have developers working in the Asia-Pacific region, but the tools they use aren’t necessary located there,” notes Gary Butler, CX1 sales manager for Cray. “The CX1 shortens the communication distance and gives companies the opportunity to decentralize their employee base.” Yet the broader impact goes far beyond the cost and availability of HPC. With rising energy costs and regulated carbon emissions, the broadened availability of HPC for high-end simulation and visualization paves the way for a comprehensive review of virtually every industrial process and the design of every manufactured product on the market. “There’s a huge push around the world for green awareness,” observes Cray’s Ungaro. “Whether it’s getting as much oil as we can out of an existing oil field, finding new ways to cool computers, or building energy-efficient autos, homes, or consumer products, all of that depends on being able to simulate how efficient and how effective these efforts will be in reality. “Ultimately, that’s what we build—machines that can simulate reality,” says Ungaro. n


To top