News You Can Use
Traffic Web Site
By Brad Moss, OD
As part of the Defense Base Realignment and Closure (BRAC) process, the National Naval Medical Center will become Walter Reed National Military Medical Center on September 15, 2011. The new Center is expected to add approximately 4,300 daily commuters and will double the number of patient visits to nearly one million annually.
The Office of Research Services and the Office of Research Facilities have developed a Web site to inform you of various construction projects and to offer tools and resources to help you with your commuting choices. The site includes timelines on construction projects, more about BRAC, and handy widgets to map the best commuter routes, complete with live traffic cameras across the entire region.
Although the increase in the number of commuters seems daunting, when viewed against the current 77,000 daily commuters in this area, it amounts to a manageable six percent increase. If NIH and others do their part to reduce traffic congestion, everyone benefits and we avoid gridlock. As employees, please consider teleworking, flexible work schedules, off-peak commuting, Transhare, and alternative commuting options such as vanpooling, carpooling, Metrorail, and NIH shuttle and public buses. Every option gets one more car off the road during peak periods. Using one or more of these strategies can make the roads accessible, reduces greenhouse gas emissions, and minimizes fuel costs.
The NIH meets regularly with state, local, and federal officials to discuss broad and specific measures to alleviate overall congestion in the area. A number of roadway improvement projects have been initiated, but these projects will create their own temporary disruptions. Even after completion, the traffic will still exceed capacity. Ultimately, it will be up to each employee to determine which mitigation strategy works best in his or her own situation. (Note: Some of these options, such as telework and alternative work schedules, require supervisor approval.)
GPU Pilot May Help Solve Computational Problems
By Andy Baxevanis, NHGRI and OIR
CIT has added a pilot set of 16 graphics processing units (GPU) nodes to the NIH Biowulf cluster to investigate new ways of solving computational problems.
GPUs are specialized microprocessors originally designed for video and rendering. More recently, computation-intensive programs in the life sciences have been ported to GPUs to explore potential performance benefits of their massive computing power.
Biowulf was designed and built at NIH and is one of the largest general-purpose biomedical computing clusters in the world. The system is designed for large numbers of simultaneous jobs that are common in bioinformatics as well as large-scale distributed memory tasks such as molecular dynamics.
During the pilot, CIT and NIH researchers will:
- Identify applications available for use with GPUs and evaluate and integrate them into the Biowulf production environment.
- Identify user simulations that will most benefit from running on GPU systems.
- Develop or port new applications to run on GPUs.
- Determine the cost and energy effectiveness of using GPU technology.
Experience shows that running any application on GPU nodes does not guarantee improved performance. It is vital that users benchmark their application to determine the effectiveness of using the GPUs. CIT staff benchmarked several biomedical applications on the GPU nodes, including NAMD, AMBER, GROMACS, MATLAB, and several bioinformatics programs.
For information on how to submit jobs to GPU nodes and to run existing GPU applications, and to see results from the application benchmarks, visit: http://biowulf.nih.gov/gpu.html. The Biowulf staff encourages users to share their GPU benchmarks by sending them to email@example.com.
One application domain found to benefit from computing with GPUs is the program NAMD for molecular dynamics simulations of large molecules (about one million atoms). With a large simulation such as STMV (satellite tobacco mosaic virus), systems with a 2:1 ratio of CPUs to GPUs give an increase in performance over that of CPU-only systems.