学術クラウド基盤mdx IIの 設計と運用 2026年6月10日 高橋 慧智（大阪大学） 

ࣗݾ঺հ w ໊લ
ߴڮ
ܛஐ
͔ͨ͸͠
͚͍ͪ
w ॴଐ
େࡕେֶ
%ηϯλʔ
εύίϯ౳ͷશࠃͷݚڀऀ͕ڞಉར༻͢Δ
ܭࢉࢿݯͷઃܭɼௐୡɼӡ༻
w ઐ໳
ߴੑೳܭࢉ
ಛʹεύίϯͷγεςϜ
ιϑτɼϓϩάϥϛϯά؀ڥɼੑೳ࠷దԽͳͲ
w ܦྺ
w େࡕେֶ
ʙ೥
w ಸྑઌ୺Պֶٕज़େֶӃେֶ
ॿڭ
ʙ೥
w ౦๺େֶαΠόʔαΠΤϯεηϯλʔ
ॿڭ
ʙ೥
w େࡕେֶ%ηϯλʔ
।ڭत
೥݄ʙ 2 

ैདྷͷ)1$γεςϜͷ՝୊ w ैདྷܕͷ)1$γεςϜ͸ɺσʔλՊֶ΍ػցֶश༻్ʹ͓͍ͯ՝୊ʹ௚໘
⿞ϋʔυ΢ΣΞ΍ιϑτ΢ΣΞͷΧελϚΠζੑ͕ݶఆత
⿞αʔόͷΑ͏ͳ௕࣮࣌ؒߦδϣϒ͸΄΅ෆՄೳ
⿞֎෦ͷσʔλݯ΍ετϨʔδͱͷ઀ଓੑ͕ශऑ 3 🤔 ιϑτʓʓΛΠϯετʔ ϧ͍͕ͨ͠SPPU͕ඞཁ ෼ੳπʔϧʓʓ͸8JOEPXTͰ͔͠ ಈ࡞͠ͳ͍ɻ 🧐 σʔληοτΛ8FC্Ͱ ެ։͍ͨ͠ ࣮ݧࢪઃ͔Β؍ଌσʔλΛ ετϦʔϛϯά͍ͨ͠ 🤨 

NEY w େֶɾݚڀػ͕ؔڞಉௐୡɾӡ༻͢ΔΫϥ΢υج൫
⿞ॊೈͳܭࢉ؀ڥͷߏங
⿞҆શͰִ཭͞ΕͨԾ૝؀ڥͷߏங
⿞σʔλιʔεͱͷγʔϜϨεͳ࿈ܞ
⿞ϦΞϧλΠϜۓٸδϣϒ΁ͷରԠ
w NEY
*͸೥ʹ౦ژେֶʹಋೖ 4 NEY
*ͷ֓ཁ
<> T. Suzumura et al., “mdx: A Cloud Platform for Supporting Data Science and Cross-Disciplinary Research Collaborations,” DASC/PiCom/CBDCom/CyberSciTech, 2022. 

NEY
**
w ࠷৽ϋʔυ΢ΣΞͷಋೖͱ஍ཧత৑௕ੑɾ଱ো֐ੑͷ࣮ݱΛ໨ࢦͯ͠೥ ݄ʹେࡕେֶʹಋೖ͞ΕͯαʔϏε։࢝
w 0QFO4UBDLΛج൫ͱͭͭ͠ɺฒྻϑΝΠϧγεςϜɺΦϒδΣΫτετϨʔ δɺϑΝΠϧϗεςΟϯά6*ɺϑΣσϨʔγϣϯೝূͳͲ༷ʑͳཁૉΛ౷߹ 5 K. Takahashi et al., “mdx II: A High-Performance Cloud Infrastructure for Data-Intensive and Cross-Disciplinary Science,” SN Computer Science, vol. 6, no.8, Dec. 2025. 

খن໛؀ڥ͔ΒεςοϓΞοϓ࣌ͷ՝୊
؅ཧίετ w Ϣʔβ͕૿͑Δʹͱ΋ͳ͍ɺཁ݅ɾཁٻ΋ ଟ༷Խ
⿞ڞ༻ΫϥελͰ͸ɺ؅ཧऀ͕ιϑτ΢Σ ΞΛ഑උ͢ΔͨΊɺશͯͷཁٻʹԠ͑Δ ͷ͸ࠔ೉
w 04ͷछྨ΍όʔδϣϯͷมߋ͸͞Βʹࠔ೉
⿞ిࢠݦඍڸͷղੳιϑτ͕8JOEPXTͰ͠ ͔ಈ͔ͳ͍
⿞όΠφϦ͔͠ఏڙ͞Εͳ͍πʔϧ͕ɺγε ςϜͷHMJCD͕ݹ͗ͯ͢ಈ͔ͳ͍
⿞ʜ 6 ϋʔυ΢ΣΞ 04
-JOVY ਺஋ܭࢉ
ϥΠϒϥϦ ίϯύΠϥ "*ϑϨʔϜ ϫʔΫ ΞϓϦέʔγϣϯ 

খن໛؀ڥ͔ΒεςοϓΞοϓ࣌ͷ՝୊
ετϨʔδੑೳ w খن໛؀ڥͰ͸/'4Ͱڞ༗ετϨʔδΛߏங͢Δ͜ͱ͕Ұൠత
⿞αʔό୆Ͱ͸ੑೳɾ༰ྔɾ଱ো֐ੑͷ޲্ʹݶք
w ϩʔΧϧετϨʔδ
ܭࢉϊʔυʹ౥ࡌ͞Ε͍ͯΔετϨʔδ
΋͋Δ͕ʜ
⿞ΞϓϦΛमਖ਼͠ɺεςʔδΠϯɾΞ΢τॲཧΛ࣮૷͢Δඞཁ
⿞େن໛σʔληοτ͸ϩʔΧϧετϨʔδʹ৐Γ੾Βͳ͍ 7 /'4αʔό /'4ΫϥΠΞϯτ TIBSF TIBSF TIBSF TIBSF σΟεΫ౥ࡌ਺ ͷݶք ΞΫηεूத 

NEY
**ͷΞʔΩςΫνϟ 8 CPUϊʔυ x60 CPU CPU NIC RAM RAM GPUϊʔυ x15 CPU CPU NIC RAM RAM GPU GPU NIC Ethernet 200Gbps GPU GPU NFS S3DS Nextcloud Object storage 432TB Lustre 553TB ؅ཧαʔό *OUFSOFU
4*/&5 

ϥοΫ֎؍ 9 (16ϊʔυ
/7*%*"
)౥ࡌ γεςϜશମ 

ܭࢉϊʔυͷߏ੒ 10 $16ϊʔυ (16ϊʔυ CPU Intel Xeon Platinum 8480+ (56 cores) x2 Intel Xeon Gold 6530 (32 cores) x2 Memory 512 GiB (DDR5-4800 SDRAM ) 1024 GiB (DDR5-5600 SDRAM ) GPU N/A NVIDIA H200 SXM5 x4 Network 200 Gbps Ethernet x1 200 Gbps Ethernet x2 # of nodes 60 15 $16ϊʔυ
ϊʔυ6 (16ϊʔυ
6 

ετϨʔδઃܭ 11 ΠϯλʔϑΣʔε ༰ྔ ໨త Block POSIX 100 TB 7.σΟεΫ Lustre POSIX 1,006 TB ߴੑೳɾฒྻ*0 S3DS S3-compatible API ߴੑೳ*0 HyperStore S3-compatible API 432 TB σʔλऩ༰ɾΞʔΧΠϒɾެ։ /dev/vda /lustre /nfs WJSUJPCML )PTU (VFTU -VTUSF /'4
4FSWFS 4%4
4FSWFS &9"4DBMFS )ZQFS4UPS F 4 4 &9"4DBMFS্ )ZQFS4UPSF্ 

Ծ૝Խج൫ͷબఆ w ઃܭஈ֊Ͱ͸W4QIFSFɺ0QFO4UBDLɺ,VCFSOFUFTͳͲΛൺֱݕ౼
w 7.8BSF
W4QIFSF͸ϥΠηϯεྉ͕ߴֹ
⿞NEYͱൺ΂Δͱ༧ࢉ͕খ͘͞ɺϊʔυ਺Λ૿΍͢͜ͱͷํ͕ॏཁͱ൑அ
w ,VCFSOFUFT͸࢖͍͜ͳͤΔϢʔβ͕ଟ͘ͳ͍
⿞૝ఆϢʔβ͸ܭࢉػͷઐ໳ՈͰ͸ͳ͍ͨΊɺௐୡ࣌఺Ͱ͸࣌ظঘૣͱ൑அ
⿞ͳ͓NEYͷޙܧػͰ͸,VCFSOFUFTΛ࠾༻༧ఆͱͷ͜ͱ 12 

0QFO4UBDLͷར఺ɾܽ఺ w Ϋϥ΢υʹظ଴͢Δ࠷௿ݶͷػೳ͸͋Δ
⿞Ծ૝ϚγϯɺετϨʔδɺωοτϫʔΫ౳
⿞࢖͍ํ΍ϊ΢ϋ΢͸Πϯλʔωοτ্ʹ๛ ෋ʹଘࡏ
w 3&45
"1*͕ॆ࣮͓ͯ͠Γ༷ʑͳπʔϧͱͷ࿈ ܞɾࣗಈԽ͕༰қ
⿞ٯʹ(6*ͰͰ͖ͳ͍ૢ࡞
Ծ૝σΟεΫͷ %-
ɺΠϝʔδͷڞ༗౳
͕͔ͳΓ͋Δ 13 

-VTUSFͷར఺ɾܽ఺ w εύίϯͳͲେن໛ΫϥελͰ޿͘༻͍Β Ε͍ͯΔฒྻ෼ࢄϑΝΠϧγεςϜ
w ༰ྔ΍ੑೳ͸΄΅ແ੍ݶɺ଱ো֐ੑ΋ߴ͍
w 104*9ޓ׵ͷͨΊɺΞϓϦͷमਖ਼͸ෆཁ
w ઃܭ΍ӡ༻͕೉͘͠ɺ044Ͱ͸͋Δ͕ϕϯ μ
%ࣾ
ͷӡ༻อक͕࣮࣭తʹෆՄܽ
w )1$؀ڥ͔Βൃల͖ͯͨͨ͠ΊɺϚϧνς φϯγͳͲΫϥ΢υ޲͚ػೳ͸΍΍ऑ͍ 14 IUUQTXJLJMVTUSFPSH
ΑΓ 

૬ޓ݁߹໢ͷઃܭ w (C&Λ࢖͍੾Δ͜ͱ͸೉͍͠
⿞Ծ૝/8Λνϡʔχϯάͯ͠΋7.ؒͷଳ Ҭ෯͸(CQTఔ౓
ෳ਺5$1઀ଓར༻
w 43*07Λ༗ޮʹ͢Δͱ(CQTୡ੒
⿞ͨͩ͠0QFO4UBDL্Ͱ43*07Λ࢖͏ͷ ͸ෆศ
࣮࣭తʹ7.͕ϊʔυʹݻఆ͞Ε Δ
w 3P$&΋)8తʹ͸ՄೳͳͨΊɺϚϧνϊʔ υͷཁ๬͕ଟ͚Ε͹ݕ౼
43*07લఏ 15 Y 

ӡ༻ͷ஌ݟ
*BB4ಛ༗ͷϢʔβαϙʔτͷ೉͠͞ w ʮࡢ೔·ͰϩάΠϯͰ͖͍ͯͨͷʹɺٸʹϩάΠϯͰ͖ͳ͘ͳͬͨʯ
⿞ϢʔβͷܭࢉϓϩάϥϜ͕ݪҼͷ00.ͩͬͨͷͰɺDHSPVQTΛ࢖ͬͨϦ ιʔε੍ݶͷํ๏ΛҊ಺
⿞εύίϯͰ͸ӡ༻ଆ͕ߦ͏޻෉͕ɺϢʔβͷ੹೚
w 7.ͷதͰ੬ऑͳιϑτ΢ΣΞͷ࣮ߦ΍ઃఆΛ͞ΕΔՄೳੑ
⿞ϑΝΠΞ΢Υʔϧ΍44)ͳͲͷجຊతͳઃఆΛҊ಺
⿞ηΩϡϦςΟϗϫΠτϖʔύʔΛެ։͠ɺ੹೚෼ք఺Λ໌֬Խ
⿞શֶͷηΩϡϦςΟ؂ࠪͷର৅ʹશ7.Λઃఆ 16 

ӡ༻ͷ஌ݟ
ॊೈੑͱੑೳͷτϨʔυΦϑ w W$16Λ෺ཧίΞʹݻఆͤͣʹӡ༻
⿞7.ͷෛՙ͸ߴ͘ͳ͍ͱ૝ఆ͠ɺࢿݯར ༻཰Λ༏ઌ
⿞7.ؒͷੑೳׯবͳͲ͕ൃੜ͠͏Δͷ Ͱɺ঎༻Ϋϥ΢υͰ͸௨ৗݻఆ
w $16ϊʔυͰ͸43*07Λෆ࢖༻
⿞ੑೳ͸্͕Δ͕ɺ7.͕/*$ͱඥ͖ͮϥΠ ϒϚΠάϨʔγϣϯ͕࢖༻ෆՄ 17 ίΞ ίΞ ίΞ ίΞ ίΞ ίΞ 7.
 7.
 7.


 W$16ݻఆ͋Γ W$16ݻఆͳ͠ ଞʹ΋ੑೳͱॊೈੑ
ࢿݯར༻ޮ཰FUD
ͷτϨʔυΦϑ͸֤ॴͰൃੜ͢Δ 

ࠓޙͷ՝୊ w (16ࢿݯͷෆ଍
⿞ϊʔυ
(16
͕طʹຒ·͓ͬͯΓधཁʹԠ͑ΒΕͳ͍ঢ়گ
⿞֦ுͷͨΊͷ༧ࢉ֫ಘʹਚྗத͕ͩɺ(16͕ߴ͍ʜ😰
w ՝ۚ؅ཧγεςϜ
⿞՝ۚ؅ཧγεςϜ͕ͳ͍ͨΊɺਓྗͰࢿݯͷ૿ݮɺ՝ֹۚͷܭࢉɺ੥ٻॻ ͷൃߦͳͲΛߦ͍ͬͯΔ
⿞εύίϯͱಉ༷ͷར༻ऀ؅ཧγεςϜΛಋೖ͠Α͏ͱ͢ΔͱɺγεςϜن໛ ʹରͯ͠ݟ߹Θͳ͍ 18