{"id":13336,"date":"2014-11-18T15:18:59","date_gmt":"2014-11-18T14:18:59","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=13336"},"modified":"2014-11-18T15:18:59","modified_gmt":"2014-11-18T14:18:59","slug":"numascale-supermicro-and-amd-announce-the-worlds-largest-shared-memory-system-to-date","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/13336-numascale-supermicro-and-amd-announce-the-worlds-largest-shared-memory-system-to-date\/","title":{"rendered":"Numascale, Supermicro, and AMD<\/span> Announce the World\u2019s Largest Shared Memory System to Date"},"content":{"rendered":"

Sys\u00adtem fea\u00adtures 5184 CPU<\/span> cores and 20.7 TBytes of shared memo\u00adry, hand\u00adles mas\u00adsi\u00adve amounts of data as the first part of a new lar\u00adge cloud com\u00adpu\u00adting faci\u00adli\u00adty for Big Data Analytics.<\/em><\/p>\n

New Orleans, Novem\u00adber 18, 2014<\/strong><\/p>\n

Numas\u00adca\u00adle today announ\u00adced the suc\u00adcessful instal\u00adla\u00adti\u00adon of a lar\u00adge shared memo\u00adry Numascale\/Supermicro\/AMD<\/span> sys\u00adtem at a cus\u00adto\u00admer dat\u00ada\u00adcen\u00adter faci\u00adli\u00adty in North Ame\u00adri\u00adca. The sys\u00adtem is the first part of a lar\u00adge cloud com\u00adpu\u00adting faci\u00adli\u00adty for ana\u00adly\u00adtics and simu\u00adla\u00adti\u00adon of sen\u00adsor data com\u00adbi\u00adned with his\u00adto\u00adri\u00adcal data.<\/p>\n

The Numas\u00adca\u00adle sys\u00adtem, instal\u00adled over the last two weeks, con\u00adsists of 108 Super\u00admi\u00adcro 1U<\/span> ser\u00advers con\u00adnec\u00adted in a 3D<\/span> torus with Numa\u00adConnect\u2122, using three cabi\u00adnets with 36 ser\u00advers apie\u00adce in a 6x6x3 topo\u00adlo\u00adgy. Each ser\u00adver has 48 cores three AMD<\/span> Opte\u00adron\u2122 6386 CPUs and 192 GBytes memo\u00adry, pro\u00advi\u00adding a sin\u00adgle sys\u00adtem image and 20.7 TBytes to all 5184 cores. The sys\u00adtem was desi\u00adgned to meet user demand for \u201cvery lar\u00adge memo\u00adry\u201d hard\u00adware solu\u00adti\u00adons run\u00adning a stan\u00addard sin\u00adgle image Linux OS<\/span> on com\u00admo\u00addi\u00adty x86 based servers.<\/p>\n

\u201c<\/span>Supermicro\u2019s 1U<\/span> 4\u2011way ser\u00advers (AS-1042G-LTF<\/span><\/a>) with Hyper\u00adTrans\u00adport (HTX<\/span>) con\u00adnec\u00adti\u00advi\u00adty pro\u00advi\u00adde users with fle\u00adxi\u00adble com\u00adpu\u00adting resour\u00adces, inclu\u00adding capa\u00adbi\u00adli\u00adties for hand\u00adling very lar\u00adge data sets such as tho\u00adse found in appli\u00adca\u00adti\u00adons for hand\u00adling near real-time sen\u00adsor inputs, com\u00adbi\u00adned with mas\u00adsi\u00adve amounts of his\u00adto\u00adri\u00adcal data,\u201d said Tau Leng, Vice Pre\u00adsi\u00addent of HPC<\/span> at Super\u00admi\u00adcro. \u201cOur 4\u2011way sys\u00adtems with can be used as a sin\u00adgle sys\u00adtem or par\u00adti\u00adtio\u00adned into smal\u00adler sys\u00adtems whe\u00adre each par\u00adti\u00adti\u00adon runs one ins\u00adtance of the OS<\/span>. With pro\u00adper Numa-awa\u00adre\u00adness, appli\u00adca\u00adti\u00adons with high band\u00adwidth requi\u00adre\u00adments will be able to uti\u00adli\u00adze the com\u00adbi\u00adned band\u00adwidth of all of the memo\u00adry con\u00adtrol\u00adlers and still be able to share data with low laten\u00adcy access through the coher\u00adent shared memory.\u201d<\/p>\n

Sys\u00adtems equip\u00adped with Numa\u00adConnect pro\u00advi\u00adde high per\u00adfor\u00admance shared memo\u00adry pro\u00adgramming capa\u00adbi\u00adli\u00adties with the same cost struc\u00adtu\u00adre as a clus\u00adter. Numa\u00adconnect\u2122 also eli\u00admi\u00adna\u00adtes the dif\u00adfi\u00adcul\u00adty of MPI<\/span> coding for big data pro\u00adblems and can increase pro\u00adgramm\u00ader pro\u00adduc\u00adti\u00advi\u00adty. This alter\u00adna\u00adti\u00adve repres\u00adents a com\u00adpel\u00adling solu\u00adti\u00adon for sci\u00aden\u00adtists who curr\u00adent\u00adly work with their shared memo\u00adry codes on x86 desk\u00adtops and lap\u00adtops. The\u00adse users can now sca\u00adle up their data sets wit\u00adhout any extra effort within a fami\u00adli\u00adar, stan\u00addard Linux OS<\/span> environment.<\/p>\n

\u201c<\/span>AMD<\/span> is proud to have been sel\u00adec\u00adted to par\u00adti\u00adci\u00adpa\u00adte in this exci\u00adting pro\u00adject,\u201d said Suresh Gopa\u00adl\u00ada\u00adkrish\u00adn\u00adan, Gene\u00adral Mana\u00adger and Cor\u00adpo\u00adra\u00adte Vice Pre\u00adsi\u00addent, Ser\u00adver Busi\u00adness Unit, AMD<\/span>.  \u201cThis col\u00adla\u00adbo\u00adra\u00adti\u00adon with Numas\u00adca\u00adle and Super\u00admi\u00adcro brings tog\u00ade\u00adther the right tech\u00adno\u00adlo\u00adgies to sol\u00adve the chal\u00adlen\u00adging pro\u00adblem of lar\u00adge shared memo\u00adry implementation.\u201d<\/p>\n

Einar Rustad, CTO<\/span> of Numas\u00adca\u00adle, agrees. \u201cThe sin\u00adgle memo\u00adry image clus\u00adter pro\u00advi\u00addes both shared memo\u00adry \u2014 inclu\u00adding threads and OpenMP \u2014 and MPI<\/span> pro\u00adgramming opti\u00adons,\u201d he said. \u201cThe sca\u00adlable sys\u00adtem takes advan\u00adta\u00adge of low cost com\u00admo\u00addi\u00adty x86 hard\u00adware to offer signi\u00adfi\u00adcant savings when com\u00adpared to con\u00adven\u00adtio\u00adnal shared memo\u00adry sys\u00adtems. With Numa\u00adConnect\u2122, sys\u00adtem admi\u00adnis\u00adtra\u00adti\u00adon is iden\u00adti\u00adcal to that of a sin\u00adgle ser\u00adver becau\u00adse the\u00adre are no sepa\u00adra\u00adte node images to main\u00adtain and dis\u00adtri\u00adbu\u00adte. We are very proud to have been part of the pro\u00adject that enab\u00adled us to build this ground\u00adbrea\u00adking sys\u00adtem in clo\u00adse coope\u00adra\u00adti\u00adon with Super\u00admi\u00adcro and AMD<\/span>.\u201d<\/p>\n

Numa\u00adConnect works with AMD<\/span> Opteron\u2122-based ser\u00advers and pro\u00advi\u00addes up to 256 TBytes of sys\u00adtem-wide shared memo\u00adry using cache cohe\u00adren\u00adcy logic with a direc\u00adto\u00adry-based pro\u00adto\u00adcol that sca\u00adles to 4096 nodes. The cache cohe\u00adren\u00adcy logic is imple\u00admen\u00adted in an ASIC<\/span> tog\u00ade\u00adther with inter\u00adcon\u00adnect fabric cir\u00adcuit\u00adry with rou\u00adting tables for mul\u00adti-dimen\u00adsio\u00adnal Torus topo\u00adlo\u00adgies. This type of fabric is very sca\u00adlable and the same topo\u00adlo\u00adgy is used in the world\u2019s lar\u00adgest supercomputers.<\/p>\n

Visit Numas\u00adca\u00adle in booth 1923 at SC14<\/span> to see live pro\u00adduct demos.<\/em><\/p>\n

About Advan\u00adced Micro Devices, Inc. (AMD<\/span>)<\/strong><\/span><\/p>\n

AMD<\/span> (NYSE<\/span>: AMD<\/span>) designs and inte\u00adgra\u00adtes tech\u00adno\u00adlo\u00adgy that powers mil\u00adli\u00adons of intel\u00adli\u00adgent devices, inclu\u00adding per\u00adso\u00adnal com\u00adpu\u00adters, tablets, game con\u00adso\u00adles, and cloud ser\u00advers that defi\u00adne the new era of sur\u00adround com\u00adpu\u00adting. AMD<\/span> solu\u00adti\u00adons enable peo\u00adp\u00adle ever\u00ady\u00adwhe\u00adre to rea\u00adli\u00adze the full poten\u00adti\u00adal of their favo\u00adri\u00adte devices and appli\u00adca\u00adti\u00adons to push the boun\u00adda\u00adries of what is pos\u00adsi\u00adble. For more infor\u00adma\u00adti\u00adon, visit www.amd.com<\/a>.<\/span><\/p>\n

About Super\u00admi\u00adcro Com\u00adpu\u00adter, Inc. <\/strong><\/span><\/p>\n

Super\u00admi\u00adcro\u00ae (NASDAQ<\/span>: SMCI<\/span>), the lea\u00adding inno\u00adva\u00adtor in high-per\u00adfor\u00admance, high-effi\u00adci\u00aden\u00adcy ser\u00adver tech\u00adno\u00adlo\u00adgy, is a pre\u00admier pro\u00advi\u00adder of advan\u00adced ser\u00adver Buil\u00adding Block Solu\u00adti\u00adons\u00ae for Data Cen\u00adter, Cloud Com\u00adpu\u00adting, Enter\u00adpri\u00adse IT<\/span>, Hadoop\/Big Data, HPC<\/span>, and Embedded Sys\u00adtems world\u00adwi\u00adde. Super\u00admi\u00adcro is com\u00admit\u00adted to pro\u00adtec\u00adting the envi\u00adron\u00adment through its \u201cWe Keep IT<\/span> Green\u00ae\u201d initia\u00adti\u00adve and pro\u00advi\u00addes cus\u00adto\u00admers with the most ener\u00adgy-effi\u00adci\u00adent, envi\u00adron\u00admen\u00adtal\u00adly-fri\u00adend\u00adly solu\u00adti\u00adons available on the mar\u00adket. For more infor\u00adma\u00adti\u00adon, visit www.supermicro.com.<\/span><\/p>\n

About Numas\u00adca\u00adle<\/strong><\/span><\/p>\n

With offices in Euro\u00adpe, Asia, and the USA<\/span>, Numascale\u2019s ground\u00adbrea\u00adking Numa\u00adConnect inter\u00adcon\u00adnect tech\u00adno\u00adlo\u00adgy is the back\u00adbone for the most sca\u00adlable ser\u00adver com\u00adpu\u00adter sys\u00adtems. Numa\u00adConnect allows high-volu\u00adme manu\u00adfac\u00adtu\u00adred ser\u00adver boards to be used as buil\u00adding blocks for sys\u00adtems with fea\u00adtures that are only found in high-end enter\u00adpri\u00adse ser\u00advers. Numa\u00adConnect includes full sup\u00adport for vir\u00adtua\u00adliza\u00adti\u00adon of pro\u00adces\u00adsing, memo\u00adry, and I\/O resour\u00adces, and can be used with stan\u00addard ope\u00adra\u00adting systems.<\/span><\/p>\n

Numas\u00adca\u00adle is sup\u00adport\u00aded by: Sta\u00adtoil, Pro\u00adVen\u00adture, Inves\u00adtinor, Inno\u00adva\u00adti\u00adon Nor\u00adway, Nor\u00adges forsk\u00adnings\u00adr\u00e5d, and Euro\u00adstars. For more infor\u00adma\u00adti\u00adon, visit www.numascale.com<\/a>.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

Sys\u00adtem fea\u00adtures 5184 CPU<\/span> cores and 20.7 TBytes of shared memo\u00adry, hand\u00adles mas\u00adsi\u00adve amounts of data as the first part of a new lar\u00adge cloud com\u00adpu\u00adting faci\u00adli\u00adty for Big Data Ana\u00adly\u00adtics. (\u2026) Wei\u00adter\u00adle\u00adsen \u00bb<\/a><\/p>\n","protected":false},"author":8,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[22],"tags":[],"class_list":["post-13336","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13336","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/comments?post=13336"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13336\/revisions"}],"predecessor-version":[{"id":13337,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13336\/revisions\/13337"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=13336"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=13336"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=13336"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}