Red Storm (computing): Difference between revisions

From Wikipedia, the free encyclopedia
Content deleted Content added
No edit summary
Avoid redirect.
 
(44 intermediate revisions by 32 users not shown)
Line 1: Line 1:
{{otheruses|Red Storm}}
{{Other uses|Red Storm (disambiguation)}}
'''Red Storm''' was a [[supercomputer]] architecture designed for the [[United States Department of Energy|US Department of Energy]]’s [[National Nuclear Security Administration]] [[Advanced Simulation and Computing Program]]. [[Cray, Inc]] developed it in 2004 based on the contracted architectural specifications provided by [[Sandia National Laboratories]].<ref>{{cite press release
|url = http://www.sandia.gov/ASC/pubs_pres/pubs/RS_flyer.pdf
|title = Red Storm 2004 fact sheet
|date = June 2004
|accessdate = 2009-08-11
|url-status = dead
|archiveurl = https://web.archive.org/web/20090511184844/http://sandia.gov/asc/pubs_pres/pubs/RS_flyer.pdf
|archivedate = 2009-05-11
}}</ref> The architecture was later commercially produced as the
[[Cray XT3]].<ref>{{cite press release
|url = http://www.sandia.gov/news-center/news-releases/2004/comp-soft-math/redstormrising.html
|title = Sandia Red Storm press release
|date = 2004-07-27
|accessdate = 2009-08-11
|url-status = dead
|archiveurl = https://web.archive.org/web/20090826103305/http://sandia.gov/news-center/news-releases/2004/comp-soft-math/redstormrising.html
|archivedate = 2009-08-26
}}</ref>


Red Storm was a partitioned, space shared, tightly coupled, [[massively parallel]] processing machine with a high performance 3D mesh network. The processors were commodity [[AMD]] [[Opteron]] CPUs with off-the-shelf memory [[DIMM]]s. The NIC/router combination, called SeaStar, was the only custom [[ASIC]] component in the system and used a [[PowerPC 400#PowerPC 440|PowerPC 440]] based core. When deployed in 2005, Red Storm’s initial configuration consisted of 10,880 single-core 2.0&nbsp;GHz Opterons, of which 10,368 were dedicated for scientific calculations. The remaining 512 Opterons were used to service the computations and also provide the user interface to the system and run a version of [[Linux]]. This initial installation consisted of 140 cabinets, taking up {{convert|280|m2}} of floor space.
'''Red Storm''' is a [[supercomputer]] architecture designed for the [[ASCI Thor's Hammer]] supercomputer at [[Sandia National Laboratory]] by [[Cray, Inc]].<ref>[http://www.sandia.gov/ASC/pubs_pres/pubs/RS_flyer.pdf Red Storm fact sheet]</ref> The architecture was later productized as the [[Cray XT3]].


The Red Storm supercomputer was designed to be highly scalable from a single cabinet to hundreds of cabinets and was scaled up twice during its lifetime. In 2006 the system was upgraded to 2.4&nbsp;GHz [[Dual-Core]] Opterons. An additional fifth row of computer cabinets were also brought online resulting in over 26,000 processor cores. This resulted in a peak performance of 124.4 teraflops, or 101.4 running the [[Linpack]] [[Benchmark (computing)|benchmark]].<ref>{{cite press release
It was announced on July 27, 2004.<ref>[http://www.sandia.gov/news-center/news-releases/2004/comp-soft-math/redstormrising.html Sandia Red Storm press release]</ref>
| url = http://www.sandia.gov/news/resources/releases/2006/red-storm.html
| title = Red Storm upgrade lifts Sandia supercomputer to 2nd in world, but 1st in scalability, say researchers
| date = 2006-11-14
| accessdate = 2009-08-11
}}</ref>
A second major upgrade in 2008 introduced [[Cray XT4]] technology: [[Quad-core]] Opteron processors and an increase in memory to 2 GB per core. This resulted in a peak theoretical performance of 284 teraflops.<ref>{{cite press release
| url = http://investors.cray.com/phoenix.zhtml?c=98390&p=irol-newsArticle&ID=1104852
| title = Cray and Sandia Announce Agreement to Upgrade "Red Storm" Supercomputer to 284 Teraflops
| date = 2008-02-06
| accessdate = 2009-08-11
}}</ref>


[[Top500]] performance ranking for Red Storm after each upgrade:
The Red Storm machine at Sandia [[Top 500]] performance:
* November 2005: Rank 6 (36.19 TFLOPS)<ref>{{Cite web
| url = http://www.top500.org/lists/2005/11
| title = Top 500 rankings for Nov 2005
| date = November 2005
| accessdate = 2009-08-11
}}</ref>
* November 2006: Rank 2 (101.4 TFLOPS)<ref>{{Cite web
| url = http://www.top500.org/lists/2006/11
| title = Top 500 rankings for Nov 2006
| date = November 2006
| accessdate = 2009-08-11
}}</ref>
* November 2008: Rank 9 (204.2 TFLOPS)<ref>{{Cite web
| url = http://www.top500.org/lists/2008/11
| title = Top 500 rankings for Nov 2008
| date = November 2008
| accessdate = 2009-08-11
}}</ref>


Red Storm was intended for capability computing. That is, a single application could be run across the entire system. This is in contrast to cluster-style capacity computing, in which portions of a cluster are assigned to run different applications. The performance of the memory subsystem, the processor, and the network must be in proper balance to achieve adequate application progress across the entire machine. System software played a key role as well. The [[Portals network programming api|Portals network programming API]] was used to ensure inter-processor communication can scale as large as the entire system, and was used on many different supercomputers, including the [[Intel Teraflops]] and [[Intel Paragon|Paragon]]. The compute processors use a custom [[lightweight kernel operating system]] named [[Catamount (operating system)|Catamount]], which was based on the operating system of [[ASCI Red]] called [[SUNMOS|"Cougar"]].<ref>{{cite press release
* November 2005: Rank 6 [http://www.top500.org/system/ranking/7653]
|url = http://www.sandia.gov/ASC/pubs-media/pubs/ASC_factsheet_RS_2008.pdf
* June 2006: Rank 9
|title = Red Storm 2008 fact sheet
* November 2006: Rank 2
|date = 2008
* June 2007: Rank 3 [http://www.top500.org/system/8193]
|accessdate = 2009-08-11
|url-status = dead
|archiveurl = https://web.archive.org/web/20090826220928/http://sandia.gov/asc/pubs-media/pubs/ASC_factsheet_RS_2008.pdf
|archivedate = 2009-08-26
}}</ref> A userspace implementation of the [[Lustre (file system)|Lustre]] file system, named [[Lustre (file system)#Implementation|liblustre]], was ported to the Catamount environment using libsysio<ref>{{cite web
| url = https://sourceforge.net/projects/libsysio/
| title = libsysio
| date = 2006
| accessdate = 2016-02-16
}}</ref> library to provide [[POSIX]]-like semantics. This filesystem client ran in the single-threaded Catamount environment without interrupts,<ref>{{cite web
| url = http://www.sandia.gov/~smkelly/SAND2006-2561C-CUG2006-CatamountDualCore.pdf
| title = Catamount Software Architecture with Dual Core Extensions
| date = 2005
| accessdate = 2016-02-16
}}</ref> and only serviced IO requests when explicitly allowed by the application, to reduce [[jitter]] introduced by background file system operations.


Red Storm was decommissioned in 2012.<ref>{{Cite web
The Top500 ranking improved between June and November of that year since the system was upgraded from 10880 2.0 GHz single core [[Opteron]]s to 12960 2.4 GHz [[dual-core]] Opterons with 26544 total processing elements for 14.5 teraflops. Additional computer cabinets were brought online as part of the "fifth-row" of the computer, bringing the total to 300 cabinets.
| url = http://www.hpcwire.com/hpcwire/2012-06-13/red_storm_passes.html

| title = Red Storm Passes
==External links and sources==
| date = June 2012
| accessdate = 2012-11-02
}}</ref>


==References==
<references/>
<references/>


{{Cray computers}}
{{Cray computers}}

{{compu-hardware-stub}}
[[Category:Cray products]]
[[Category:Parallel computing]]
[[Category:X86 supercomputers]]
[[Category:United States Department of Energy]]

Latest revision as of 08:22, 1 May 2024

Red Storm was a supercomputer architecture designed for the US Department of Energy’s National Nuclear Security Administration Advanced Simulation and Computing Program. Cray, Inc developed it in 2004 based on the contracted architectural specifications provided by Sandia National Laboratories.[1] The architecture was later commercially produced as the Cray XT3.[2]

Red Storm was a partitioned, space shared, tightly coupled, massively parallel processing machine with a high performance 3D mesh network. The processors were commodity AMD Opteron CPUs with off-the-shelf memory DIMMs. The NIC/router combination, called SeaStar, was the only custom ASIC component in the system and used a PowerPC 440 based core. When deployed in 2005, Red Storm’s initial configuration consisted of 10,880 single-core 2.0 GHz Opterons, of which 10,368 were dedicated for scientific calculations. The remaining 512 Opterons were used to service the computations and also provide the user interface to the system and run a version of Linux. This initial installation consisted of 140 cabinets, taking up 280 square metres (3,000 sq ft) of floor space.

The Red Storm supercomputer was designed to be highly scalable from a single cabinet to hundreds of cabinets and was scaled up twice during its lifetime. In 2006 the system was upgraded to 2.4 GHz Dual-Core Opterons. An additional fifth row of computer cabinets were also brought online resulting in over 26,000 processor cores. This resulted in a peak performance of 124.4 teraflops, or 101.4 running the Linpack benchmark.[3] A second major upgrade in 2008 introduced Cray XT4 technology: Quad-core Opteron processors and an increase in memory to 2 GB per core. This resulted in a peak theoretical performance of 284 teraflops.[4]

Top500 performance ranking for Red Storm after each upgrade:

  • November 2005: Rank 6 (36.19 TFLOPS)[5]
  • November 2006: Rank 2 (101.4 TFLOPS)[6]
  • November 2008: Rank 9 (204.2 TFLOPS)[7]

Red Storm was intended for capability computing. That is, a single application could be run across the entire system. This is in contrast to cluster-style capacity computing, in which portions of a cluster are assigned to run different applications. The performance of the memory subsystem, the processor, and the network must be in proper balance to achieve adequate application progress across the entire machine. System software played a key role as well. The Portals network programming API was used to ensure inter-processor communication can scale as large as the entire system, and was used on many different supercomputers, including the Intel Teraflops and Paragon. The compute processors use a custom lightweight kernel operating system named Catamount, which was based on the operating system of ASCI Red called "Cougar".[8] A userspace implementation of the Lustre file system, named liblustre, was ported to the Catamount environment using libsysio[9] library to provide POSIX-like semantics. This filesystem client ran in the single-threaded Catamount environment without interrupts,[10] and only serviced IO requests when explicitly allowed by the application, to reduce jitter introduced by background file system operations.

Red Storm was decommissioned in 2012.[11]

References[edit]

  1. ^ "Red Storm 2004 fact sheet" (PDF) (Press release). June 2004. Archived from the original (PDF) on 2009-05-11. Retrieved 2009-08-11.
  2. ^ "Sandia Red Storm press release" (Press release). 2004-07-27. Archived from the original on 2009-08-26. Retrieved 2009-08-11.
  3. ^ "Red Storm upgrade lifts Sandia supercomputer to 2nd in world, but 1st in scalability, say researchers" (Press release). 2006-11-14. Retrieved 2009-08-11.
  4. ^ "Cray and Sandia Announce Agreement to Upgrade "Red Storm" Supercomputer to 284 Teraflops" (Press release). 2008-02-06. Retrieved 2009-08-11.
  5. ^ "Top 500 rankings for Nov 2005". November 2005. Retrieved 2009-08-11.
  6. ^ "Top 500 rankings for Nov 2006". November 2006. Retrieved 2009-08-11.
  7. ^ "Top 500 rankings for Nov 2008". November 2008. Retrieved 2009-08-11.
  8. ^ "Red Storm 2008 fact sheet" (PDF) (Press release). 2008. Archived from the original (PDF) on 2009-08-26. Retrieved 2009-08-11.
  9. ^ "libsysio". 2006. Retrieved 2016-02-16.
  10. ^ "Catamount Software Architecture with Dual Core Extensions" (PDF). 2005. Retrieved 2016-02-16.
  11. ^ "Red Storm Passes". June 2012. Retrieved 2012-11-02.