Xav tau lub tshuab uas tuaj yeem ua trillions ntawm ntab-ntsiab lus suav hauv ib pliag? Puas yog koj xav tau zaj dab neeg txias txog koj tus kheej lub supercomputer uas tua lub teeb hauv koj lub zos? Tsim lub supercomputer yog qhov nyuaj rau koj yog tias koj yog tus muaj peev xwm ntse nrog qee lub sijhawm. Kev siv tshuab, tshuab raj ntau lub tshuab yog lub network ntawm cov khoos phis tawj uas ua haujlwm ua ke los daws teeb meem. Kab lus no yuav tham luv luv txog txhua theem ntawm nws kev tsim, tsom mus rau kho vajtse thiab software.
Kauj ruam
Kauj Ruam 1. Ua ntej, nrhiav seb yam khoom kho vajtse twg koj xav tau
Ib lub hauv paus tseem ceeb, tsawg kawg yog kaum ob tus lej sib xws, hloov Ethernet, chav faib hluav taws xob (PDU), thiab khib server. Kuj nrhiav paub txog hluav taws xob, cua txias, thiab qhov chaw xav tau. Qhia meej IP chaw nyob rau tus kheej network, cov npe ntawm cov node, pob software uas koj xav tau teeb tsa, thiab thev naus laus zis twg koj xav siv los ua kom lawv txhua tus ua haujlwm ua ke kom ua haujlwm sib npaug (ntau ntxiv hauv qab no).
- Txawm hais tias kho vajtse koj xav tau yog kim, software hauv phau ntawv qhia no yog txhua yam pub dawb, thiab feem ntau nws yog qhov qhib.
- Yog tias koj xav pom sai npaum li cas koj lub supercomputer yuav nrawm (hauv kev xav), siv qhov no:
Kauj Ruam 2. Tsim cov lej
Koj tuaj yeem sib sau ua ke cov lej uas koj xav tau koj tus kheej lossis siv lub server npaj tau.
- Xaiv lub koos pij tawj lub server uas ua kom muaj chaw ntau, txias, thiab ua haujlwm zoo.
- Lossis koj tuaj yeem siv kwv yees li kaum ob tus neeg rau zaub mov qub qub. Thaum siv ua ke, lawv muaj txiaj ntsig ntau dua li thaum siv tus kheej, thiab koj tuaj yeem txuag tau me ntsis. Txhua tus txheej txheem, kev hloov pauv hauv network, thiab motherboards yuav tsum zoo ib yam kom ntseeg tau tias lub kaw lus ua haujlwm tau zoo. Tau kawg, tsis txhob hnov qab RAM thiab lub peev xwm khaws cia rau txhua qhov thiab tsawg kawg yog ib qho kho qhov muag rau lub hauv paus loj.
Kauj Ruam 3. Mount tus neeg rau zaub mov uas koj tau tsim rau hauv lub server
Pib hauv qab kom tsis txhob muaj kev tawm tsam rau saum. Caw ib tus phooj ywg los pab koj, vim tias cov neeg rau zaub mov tsis sib xws tuaj yeem hnyav heev uas nws ua rau nws tsis yooj yim kom haum rau hauv cov tub rau khoom.
Kauj Ruam 4. Mount Ethernet hloov rau saum tus ncej server
Siv lub sijhawm no los teeb tsa nws: muab nws lub thav duab loj ntawm 9000 bytes, teeb tus IP chaw nyob rau qhov chaw nyob zoo li koj tau teev tseg hauv qib 1, thiab tua cov txheej txheem tsis tseem ceeb xws li SMTP Snooping.
Kauj Ruam 5. Txhim kho chav faib hluav taws xob
Koj yuav xav tau 220 volts rau kev ua haujlwm siab, nyob ntawm seb muaj pes tsawg tam sim no ntawm qhov xav tau ntawm qhov siab tshaj plaws.
Kauj Ruam 6. Thaum txhua yam raug teeb tsa, koj tuaj yeem pib txheej txheem teeb tsa
Linux yog qhov yuav tsum tau siv ua haujlwm rau kev ua haujlwm tau zoo hauv pawg, vim tias sib nrug los ntawm qhov zoo tshaj plaws rau kev suav suav, nws tseem yog 100% pub dawb. Nrog cov node uas mus txog ntau pua lossis ntau txhiab leej, nws yuav twv yuav raug nqi heev yog tias koj siv Windows!
- Pib los ntawm kev txhim kho qhov tseeb version ntawm motherboard BIOS thiab firmware. Lub khoos phis tawj yuav tsum zoo ib yam rau txhua qhov. Pib nrog kev txhim kho qhov tseeb version ntawm motherboard BIOS thiab firmware, uas yuav tsum zoo ib yam ntawm txhua qhov.
- Nruab Linux distro koj xav tau ntawm txhua lub node, nrog cov duab sib cuam tshuam ntawm lub hauv paus loj. Kev xaiv nrov yog CentOS, OpenSuse, Scientific Linux, RedHat, thiab SLES.
- Tus sau xav kom siv Rocks Cluster Distribution. Pob Zeb yuav teeb tsa tag nrho cov haujlwm uas koj lub supercomputer xav tau los ua haujlwm, thiab siv txoj hauv kev zoo los 'qhia' nws tus kheej hla txhua qhov ntawm tam sim no uas siv Red Hat's PXE boot thiab 'Kick Start' txheej txheem.
Kauj Ruam 7. Txhim kho qhov sib tham sib tham, tswj cov peev txheej, thiab lwm cov software qiv qiv tseem ceeb
Yog tias koj tsis tau teeb tsa Rocks hauv cov kauj ruam dhau los, koj yuav tsum tau npaj cov software xav tau los ua kom muaj zog sib npaug suav nrog koj tus kheej.
- Ua ntej tshaj, koj yuav xav tau lub tshuab tswj hwm kev txav mus los zoo li Torque Tus Thawj Saib Xyuas Haujlwm, uas yuav ua txoj haujlwm sib koom ntawm cov tshuab.
- Pair Torque nrog Maui Cluster Scheduler kom ua tiav kev teeb tsa.
- Tom ntej no, koj yuav tsum teeb tsa lub ntsiab lus sib tham, uas xav tau los ua cov lej sib cais sib faib cov ntaub ntawv tib yam. OpenMP yog qhov kev xaiv meej.
- Tsis txhob hnov qab cov tsev qiv ntawv ua lej ntau xov thiab cov neeg sib sau los tsim cov phiaj xwm sib dhos uas koj xav tau. Lossis, tsuas yog nruab Rocks kom yooj yim dua.
Kauj Ruam 8. Muab tag nrho cov lej suav rau hauv lub network
Lub hauv paus tseem ceeb yuav xa cov haujlwm ua haujlwm mus rau cov lej suav, uas tom qab ntawd yuav tsum xa cov txiaj ntsig rov qab thaum sib tham sib tham. Qhov sai dua qhov zoo dua.
- Siv lub network ethernet ntiag tug los txuas txhua qhov ntawm koj pawg supercomputer.
- Thawj qhov tuaj yeem yog NFS, PXE, DHCP, TFTP, thiab NTP server hauv ethernet network.
- Koj yuav tsum cais lub network no los ntawm cov pej xeem network kom ntseeg tau tias cov pob ntawv xa mus tsis cuam tshuam nrog lwm lub network hauv koj lub network.
Kauj Ruam 9. Kuaj lub supercomputer uas koj tau tsim
Ua ntej siv los ntawm lwm tus, peb pom zoo kom koj sim ua haujlwm ntawm koj lub supercomputer ua ntej. HPL (Kev Ua Tau Zoo Linpack) yog qhov ntsuas tau zoo rau ntsuas kev suav lub nrawm ntawm cov kws siv tshuab supercomputers. Koj yuav tsum tau suav sau los ntawm qhov chaw, nrog rau txhua qhov kev xaiv ua kom zoo uas muab los ntawm tus neeg sau uas koj siv rau cov vaj tsev uas koj tau xaiv.
- Tau kawg, koj yuav tsum tau suav sau los ntawm qhov chaw nrog txhua qhov ua tau zoo xaiv rau koj lub platform. Piv txwv, yog tias siv AMD CPU, suav nws siv Open64 nrog qib kev ua kom zoo -0 nrawm.
- Sib piv koj cov txiaj ntsig ntsuas ntawm TOP500.org txhawm rau sib piv koj lub supercomputer nrog 500 lub supercomputers nrawm tshaj plaws hauv ntiaj teb!
Lub tswv yim
- Rau siab network nrawm, ua tibzoo saib InfiniBand network interface. Tau kawg, koj yuav tsum tau npaj them tus nqi hwm.
- IPMI tuaj yeem ua kom yooj yim rau kev tswj hwm ntawm cov supercomputer loj los ntawm kev muab KVM-dhau-IP, kev tswj hwm lub zog hluav taws xob nyob deb, thiab lwm yam.
- Siv Ganglia los saib xyuas kev suav cov load ntawm cov ntshav.
