Real-world lessons from a HPC system deployment  

• Learn about issues and challenges in the lifecycle of an academic cluster
• Managing communications and expectations
• Challenges of integrating hardware and software
• Collaborating with vendor to isolate and remedy issues
• What worked, what didn’t
• Assessing user experience and barriers to adoption of technology
• Applying lessons learned to future deployments
• Lively discussion on strategies for meeting challenges to follow
  

In an ideal world, High Performance Computing (HPC) deployments are well-planned projects that rapidly deliver systems to meet user needs according to pre-determined plans. This presentation reviews the deployment of a small HPC system at the University of Alabama (UA), specifically focusing on lessons learned about the design, deployment, operation and support of a cluster system in our academic environment. The goal of the presentation is to highlight specific challenges we faced in our deployment and the varying levels of success we have experienced in managing those challenges. The presenter intends to stimulate a sharing of challenges and strategies for meeting those challenges.

The discussion will cover the life cycle of our system (design, implementation, production, support and future planning). The primary issues in the design phase of the UA HPC cluster were related to communication and expectation and how they affected the ensuing stages of the deployment and operation of the cluster.

The discussion will then move to an overview of the cluster design and a review of the challenges of integrating the hardware and software during the implementation stage. We will also discuss how we worked with the vendor to reconcile the different expectations of the actual system to be built. In the discussion on the production, we will highlight design decisions that didn't produce the expected value, as well as aspects of the system that outperformed our expectations.

The presentation will also review the user experience and user uptake of the HPC technology and covers some of the technical, political and personal barriers that can prevent full utilization of a central HPC system. Finally, with the benefit of experience, we speculate on how to apply our lessons learned to future system deployments so we can deliver better solutions for our researchers and reduce administrative overhead.

Exploring the effect of HPC on biomedical sciences   

• Meet the challenges of industry-specific job scheduling needs
• Learn best practices of deploying shared computational resources
• Discover ideas on technical and process blueprints for HPC implementations
• Explore a working shared resource infrastructure based on Platform LSF

In recent years, High Performance Computing (HPC) has moved from the sidelines to the mainstream of biomedical research. Increasingly researchers are employing computational methods to facilitate their wet lab research. In fact, some emerging laboratories and approaches are based on a 100% computational framework. While there are many lessons to be learned from the computational infrastructure put into place for the physical and mechanical sciences, the character, nature and demands of biomedical computing differ from the needs of the other sciences. Biomedical computational problems, for example, tend to be less computationally intensive but more "bursty". This creates both an opportunity (it is easier to meet capacity needs) and a challenge (job scheduling rules are more complicated to accommodate the bursts).

In 2007, Harvard convened the first Biomedical High Performance Computing Leadership Summit to explore the issues in creating shared computing infrastructure for the biomedical sciences. We brought together over 100 leaders in the field to exchange ideas and approaches. Through special sessions and direct participant surveys a number of themes emerged around best practices in deploying shared computational infrastructure for the biomedical sciences. In this presentation we summarize the approaches and ideas to provide a technical and process blueprint for organizations wishing to provide shared research computing research resources for groups small or large - from a few hundred CPUs to thousands of CPUs. In particular we highlight the use of Platform LSF to create shared infrastructure that works.

Tapan Patel, Product Marketing Manager, SAS - Bio
Carl Anderson, Office of the CTO, IBM Systems and Technology - Bio
Karl Cain, Business Development Manager, Dell - Bio
Christian Belady, Principal Power and Cooling Architect, Global Foundation Services Group, Microsoft - Bio
Bob Moore, Group Manager, Marketing Strategy Team, Industry Standard Server Division, HP - Bio

• Hear varying perspectives on what green really means
• Participate in a lively discussion on the viability of the green message in HPC
• Find out what customers are doing to save energy for power and cooling
• Learn about the indirect green benefits of HPC

As High Power Computing (HPC) datacenters grow ever larger, so does the amount of electricity used to power and cool the powerful servers. With energy shortages and energy costs ranking at the top of environmental concerns, a movement to decrease HPC datacenter energy usage is gathering momentum. 

A particular concern is that power consumption is rising faster than the average datacenter can accommodate.  It’s not just a matter of building out of space to house new servers. IT also needs to add additional electrical capacity and cooling systems. Gartner analysts advise enterprises to make the most efficient use of existing systems with technologies such as workload balancing and virtualization, and to plan for greener datacenter designs in the years ahead.

A panel comprised of industry leaders including IBM, HP, Dell, SAS and Platform will tackle these challenges and discuss the innovative strategies their companies are undertaking to help organizations create greener HPC datacenters. Participants will have an opportunity to participate by asking questions of the panel that will help them design their own beneficial green strategy.

Get It Faster, Run IT Smarter, Grow IT Better: In the ongoing race towards efficiencies; cost, performance, power, storage coupled with the changing landscapes of “systems architectures” and support technologies IT management is faced with more and growing complexity.

As the shift to pure commodity based components becomes today’s architectural mainstay, the challenge then to technology providers is to continue to deliver solutions where the “sum of parts” are inconsequential to both the datacenter operations team as well as to their customer base.

Today, IT should … just work!

This talk will cover Dell’s leadership in the charge towards “standardization” of cluster based architectures, its leadership in “Simplifying IT” and the benefits that support customers desire for ease of everything –fostered by Dell’s long term partnership with Platform Computing and its suite of software designed to help in the rapid deployment of scale-out architectures and Grid-based solutions.

As the world-wide leader in High Performance Computing, HP delivers HPC solutions that enable it’s customers to focus on their engineering, science and analytical goals, allowing them to analyze large amounts of data and perform increasingly complex simulations, providing greater insight and helping them to make better decisions. To do that, HP maintains a team of HPC experts who act as trusted advisors to HP’s customers, utilizing HP’s HPC computing environment as a testbed for customer problems. Just as its customers have deployed Grid technology to better enable them to meet their goals, HP has implemented a global grid to share its world-wide HPC benchmarking resources.

Interoperability and Mixed OS Grid Environments

Panel members will discuss their experience with mixed operating system clusters.  The discussion will include their objectives and challenges, both on the business and technical fronts, of introducing and maintaining mixed grids.  The panel includes representatives from industry segments in electronics, finance, Global ISV, and the public sector.

High Performance Computing in Materials Science

This presentation will be focusing on applications that simulate catatrosphic failure of metals, composites and biological tissues.
Outlined will be the historical development and the impact of HPC in the materials science field, discussion of problems that
have been addressed and overview of future avenues and opportunities.

Also presented will be an overview of multi-scale simulation challenges and applications to biological materials and biological structures,
a field that emerges as oneof the critical frntiers in the field for which computation is believed to be a key player. The usefulness of
computations models to advance science and/or our ability to engineer systems, materials or structures at scales from nano to macro will
also be discussed. This will be crucial for future applications and funding, which will be measured by the impact generated from our community.

Usage of Platform Analytics and Platform RTM in the Automotive Industry

The automotive development is using more and more batch functionality, not only in the known CAE HPC market, this usage also become more and more important in the product development and the data management (PDM)

Automotive OEM’s worldwide have one common goal - They would get a better services for a better price, additional the reporting requirements regarding Utilization, Throughput and Accounting information increases.

This presentation should show how Platform Analytics in combination with LSF RTM could help to achieve this goals.