In die­ser Sek­ti­on wer­den Pres­se­mit­tei­lun­gen der Her­stel­ler 1:1 ver­öf­fent­licht. Pla­net 3DNow! über­nimmt weder für die Rich­tig­keit der Anga­ben noch für die Art der For­mu­lie­rung jed­we­de Ver­ant­wor­tung. Die Anga­ben kön­nen sub­jek­tiv sein und geben nicht not­wen­di­ger­wei­se die Mei­nung der auf Pla­net 3DNow! schrei­ben­den Autoren wieder.

CRAY INTRODUCES CLUSTERSTOR E1000 STORAGE TO FUEL CONVERGED WORKLOADS OF THE EXASCALE ERA

Verfasst vonpipin
 
New HPC Par­al­lel Sto­rage Sys­tem Deli­vers Mas­si­ve Sca­la­bi­li­ty and Unpre­ce­den­ted Per­for­mance to Power Data-dri­ven Workloads Inclu­ding AI, Ana­ly­tics, Simu­la­ti­on and Modeling

SEATTLE, Oct. 30, 2019 (GLOBE NEWSWIRE) — Glo­bal super­com­pu­ter lea­der Cray, a Hew­lett Packard Enter­pri­se com­pa­ny (NYSE: HPE), today unvei­led its Cray Clus­terS­tor E1000 sys­tem, an enti­re­ly new par­al­lel sto­rage plat­form for the Exas­ca­le Era. Clus­terS­tor E1000 addres­ses the explo­si­ve growth of data from con­ver­ged workloads and the need to access that data at unpre­ce­den­ted speed, by offe­ring an opti­mal balan­ce of sto­rage per­for­mance, effi­ci­en­cy and sca­la­bi­li­ty, effec­tively eli­mi­na­ting job pipe­line con­ges­ti­on cau­sed by I/O bot­t­len­ecks. The next-gene­ra­ti­on glo­bal file sto­rage sys­tem has alre­a­dy been sel­ec­ted by the US Depart­ment of Ener­gy (DOE) for use at the Argon­ne Lea­der­ship Com­pu­ting Faci­li­ty, Oak Ridge Natio­nal Labo­ra­to­ry and Law­rence Liver­mo­re Natio­nal Labo­ra­to­ry, whe­re the first three US exas­ca­le super­com­pu­ters will be housed (respec­tively Auro­ra, Fron­tier and El Capi­tan). With the intro­duc­tion of the Clus­terS­tor E1000 sto­rage sys­tem, Cray has com­ple­ted the re-archi­tec­tu­re of its end-to-end infra­struc­tu­re port­fo­lio, which encom­pas­ses Cray Shas­ta super­com­pu­ters, Cray Slingshot inter­con­nect, and the Cray soft­ware plat­form. With Cray’s next-gene­ra­ti­on end-to-end super­com­pu­ting archi­tec­tu­re, available for any dat­a­cen­ter envi­ron­ment, cus­to­mers around the world can unleash the full poten­ti­al of their data.

“To hand­le the mas­si­ve growth in data that cor­po­ra­ti­ons world­wi­de are deal­ing with in their digi­tal trans­for­ma­ti­ons, a com­ple­te­ly new approach to sto­rage is requi­red,” said Peter Unga­ro, pre­si­dent and CEO of Cray, a Hew­lett Packard Enter­pri­se com­pa­ny. “Cray’s new sto­rage plat­form is a com­pre­hen­si­ve rethin­king of what high per­for­mance sto­rage means for the Exas­ca­le Era. The intel­li­gent soft­ware and hard­ware design of Clus­terS­tor E1000 orchestra­tes the data flow with the work­flow – that’s some­thing no other solu­ti­on on the mar­ket can do.”

As the exter­nal high per­for­mance sto­rage sys­tem for the first three U.S. exas­ca­le sys­tems, Cray Clus­terS­tor E1000 will total over 1.3 exa­bytes of sto­rage for all three sys­tems com­bi­ned. The Natio­nal Ener­gy Rese­arch Sci­en­ti­fic Com­pu­ting Cen­ter (NERSC) also sel­ec­ted Clus­terS­tor E1000, which will be the industry’s first all NVMe par­al­lel file sys­tem at a sca­le of 30 peta­bytes of usable capacity.

“NERSC will deploy the new Clus­terS­tor E1000 on Perl­mut­ter as our fast all flash sto­rage tier, which will be capa­ble of over four tera­bytes per second wri­te band­width. This archi­tec­tu­re will sup­port our diver­se workloads and rese­arch disci­pli­nes,” said NERSC Direc­tor Sudip Dosan­jh. “Becau­se this file sys­tem will be the first all-NVMe file sys­tem deploy­ed at a sca­le of 30 peta­bytes usable capa­ci­ty, exten­si­ve quan­ti­ta­ti­ve ana­ly­sis was under­ta­ken by NERSC to deter­mi­ne the opti­mal archi­tec­tu­re to sup­port the work­flows our rese­ar­chers and sci­en­tists use across bio­lo­gy, envi­ron­ment, che­mis­try, nuclear phy­sics, fusi­on ener­gy, plas­ma phy­sics and com­pu­ting research.”

Reco­gni­zing the data access chal­lenges pre­sen­ted by the Exas­ca­le Era, Cray’s Clus­terS­tor E1000 enables orga­niza­ti­ons to achie­ve their rese­arch mis­si­ons and busi­ness objec­ti­ves fas­ter by offering:

  • Unpre­ce­den­ted sto­rage per­for­mance: Clus­terS­tor E1000 sys­tems can deli­ver up to 1.6 tera­bytes per second and up to 50 mil­li­on I/O ope­ra­ti­ons per second per rack – more than dou­ble com­pared to other par­al­lel sto­rage sys­tems in the mar­ket today.
  • Maxi­mum per­for­mance effi­ci­en­cy: New pur­po­se-engi­nee­red end-to-end PCIe 4.0 sto­rage con­trol­lers ser­ve the maxi­mum per­for­mance of the under­ly­ing sto­rage media to the com­pu­te nodes and new intel­li­gent Cray soft­ware, Clus­terS­tor Data Ser­vices, allows cus­to­mers to ali­gn the data flow with their spe­ci­fic work­flow, mea­ning they can place the appli­ca­ti­on data at the right time on the right sto­rage media (SSD pool or HDD pool) in the file system.
  • Mas­si­ve sca­la­bi­li­ty: An ent­ry-level sys­tem starts at 30 giga­bytes per second and at less than 60 tera­bytes usable capa­ci­ty. Cus­to­mers can start at the size dic­ta­ted by their cur­rent needs and sca­le as tho­se needs grow, with maxi­mum archi­tec­tu­ral head­room for future growth. The Clus­terS­tor E1000 sto­rage sys­tem can con­nect to any HPC com­pu­te sys­tem that sup­ports high speed net­works like 200 Gbps Cray Slingshot, Infi­ni­band EDR/HDR and 100/200 Gbps Ethernet.

“Cray’s Shas­ta archi­tec­tu­re sub­stan­ti­al­ly expands the company’s addressa­ble mar­ket to include HPC simu­la­ti­on, AI, enter­pri­se ana­ly­tics and cloud com­pu­ting. Clus­terS­tor E1000 is an inte­gral part of this stra­tegy,” said Ste­ve Con­way, seni­or vice pre­si­dent of rese­arch at Hype­ri­on Rese­arch. “This new edi­ti­on of the pro­ven Clus­terS­tor solu­ti­on is desi­gned to enable lea­ding enter­pri­ses to con­so­li­da­te their AI, HPC and High Per­for­mance Data Ana­ly­sis stacks, effi­ci­ent­ly and easi­ly. The fast-gro­wing con­tin­gent of enter­pri­ses that are adop­ting HPC now have the cost-effec­ti­ve opti­on to acqui­re a uni­fied Cray Shas­ta-Slingshot-Clus­terS­tor infrastructure.”

Cray Clus­terS­tor E1000 sys­tems will be available start­ing in Q1 of 2020. For more infor­ma­ti­on, plea­se cont­act your aut­ho­ri­zed Cray sales repre­sen­ta­ti­ve or visit www.cray.com.

About Cray Inc.
Cray, a Hew­lett Packard Enter­pri­se com­pa­ny, com­bi­nes com­pu­ta­ti­on and crea­ti­vi­ty so visio­na­ries can keep asking ques­ti­ons that chall­enge the limits of pos­si­bi­li­ty. Dra­wing on more than 45 years of expe­ri­ence, Cray deve­lo­ps the world’s most advan­ced super­com­pu­ters, pushing the boun­da­ries of per­for­mance, effi­ci­en­cy and sca­la­bi­li­ty. Cray con­ti­nues to inno­va­te today at the con­ver­gence of data and dis­co­very, offe­ring a com­pre­hen­si­ve port­fo­lio of super­com­pu­ters, high-per­for­mance sto­rage, data ana­ly­tics and arti­fi­ci­al intel­li­gence solu­ti­ons. Go to www.cray.com for more information.