SGI Releases New HPC Blade Platform



By Allison Proffitt

August 8, 2007 |  SGI’s recently announced Altix ICE product is positioned to bridge the divide between the high-throughput and high-performance environments. ICE is based on the Intel 5000X chipset and features 32GB of memory per blade, with up to 512 Intel Xeon processor cores per rack delivering 6 TFLOPS of performance.

“Customers are tired of clusters,” SGI sciences segment manager, Michael Brown, told Bio•IT World, listing problems such as productivity disappointments and struggles powering and cooling clusters. But SGI thinks the Altix ICE solution will alleviate many of those headaches.

ICE relies on several technologies to offer what SGI promises is breakthrough reliability, density, simplicity, power, and cooling options. SGI expects these advances to make ICE a logical solution for life science tasks requiring both multiple CPU cores per job and multiple parallel jobs.

ICE is n+1 configured; power and cooling backups are redundant, but conserved. If a power supply or fan fails, the system is hot-swapped to the remaining supplies until the problem is fixed. Paired with SGI’s “ESP” software that contacts SGI when a feature malfunctions, the configuration is reliable and efficient. “We’ve had SGI call [an ICE early user] about a fan failure, and the site didn’t even know there was a problem,” said Brown. The fixes were made without any system interruption.

Using SGI Altix 4700 innovations, the system runs at 76% energy efficiency at the rack level, and is cooled with a multi-door water cooling system. Especially for smaller bioscience operations, said Brown, the dense configuration of each rack and the efficiency of the cooling system will allow organizations to pack more racks into smaller space.

Finally, Brown touted the system’s scalability and performance. Operating system synchronization is integrated across nodes. A 2% increase in overhead efficiency, he said, can lead to a three-fold increase in computing power by conserving wasted cycles. In addition, InfiniBand connected storage is faster than local disk storage. The result  is a system that functions “much closer to a single system than clusters,” for less cost.

The system connects with a duel InfiniBand backplane, and can be installed and running in two hours, Brown said.

Subscribe to Bio-IT World  magazine.

Click here to login and leave a comment.  

0 Comments

Add Comment

Text Only 2000 character limit

Page 1 of 1



White Papers & Special Reports

sgi whp 2
Managing the Modern Genomics Data Flood
Sponsored by SGI

Managing and storing the perfect storm of multi-disciplined data pouring from next generation sequencers and other omics instruments is a central challenge in life sciences. Discover in this paper how the SGI ArcFiniti storage solution, optimized for unstructured genomics and life sciences data can: 

  • Reduce costs, proactively protect data integrity, and deliver the high performance I/O required for genomics data processing and analysis.  
  • Effectively manage capacities from 156TB to 1.4PB as a disk based, integrated hardware and software platform 


sgi - whp 1
Turning Genomics Data into Practical Insight
Sponsored by SGI

With worldwide sequencing capacity approaching 13 quadrillion DNA bases annually turning genomics data into knowledge is a true computational challenge. Read this paper and learn how the SGI UV coherent shared memory platform can:  

  • Speed results time while cost competitively tackling the most difficult computational problems across all omics disciplines. 
  • Push performance by scaling to extraordinary levels, up to 256 sockets (2,560 cores, 4,096 threads) per single system (one OS image). 

Provide support for up to 16TB of coherent shared memory in a single system image enabling extreme efficiency across a wide range of compute demands. 



accerlys-logo_2012_wh
New Complimentary Market Survey…
Collaborations and Communications Within Drug Discovery Research
Sponsored by Accelrys
This survey was conducted by the Cambridge Healthtech Media Group in January, 2012. It was sponsored by Accelrys related to their HEOS initiative to gather valid information around externalizing collaborative research while improving communications in the cloud. With 310 qualified industry respondents the survey findings reveal useful usage and trends patterns.  An insightful follow-on discussion and webinar related to this survey, and the HEOS by Scynexis SaaS portal is also available on the Bio-IT World website for complementary viewing.
 


Job Openings

tessella logo 
Scientific Software Engineer
Boston MA
$70,000 to $95,000
 
Apply at http://jobs.tessella.com   

oxford nanopore logo 


Early Access Collaborations ManagersClick here to find out more and apply   

Oxford Nanopore's GridION technology, VP, Sales and Marketing Click to  Apply  

For reprints and/or copyright permission, please contact  Tim McLucas, (781) 972-1342, tmclucas@healthtech.com .