[論文紹介] TiDB:a Raft-based HTAP database
今回ã¯TiDB（PingCAP） ä¹…ã—ã¶ã‚Šã«è«–文紹介シリーズã®ç¬¬2å¼¾ã§ã‚る。 今回ã¯åˆ†æ•£DBã®ã©çœŸã‚“ä¸ã€PingCAPãŒé–‹ç™ºã—ã¦ã„ã‚‹TiDBã«ã¤ã„ã¦ã®è«–文「TiDB:a Raft-based HTAP databaseã€(VLDB2020)を紹介ã™ã‚‹ã€‚ ã“ã®è«–æ–‡ã«é–¢ã—ã¦ã¯ã€PingCAP社ãŒè‡ªèº«ã®ãƒ–ãƒã‚°ã§ã‚‚解説ã—ã¦ã„る。 TiDBã¨ã¯ 念ã®ãŸã‚ã€TiDBã¨ã¯ä½•ãªã®ã‹ã‚’触れã¦ãŠã“ã†ã€‚ 一言ã§ã„ã†ã¨ã€ã€ŒMySQL互æ›ã®NewSQL（ï¼åˆ†æ•£SQLデータベース）ã€ã§ã‚る。 NewSQLã¨ã¯ä½•ã‹ã«ã¤ã„ã¦ã®èª¬æ˜Žã¯ä»Šå›žè¨˜äº‹ã§ã¯çœç•¥ã™ã‚‹ãŒã€éŽåŽ»ã«æ›¸ã„ãŸã“ã¡ã‚‰ã®å…¥é–€ç·¨ã‚„ã“ã¡ã‚‰ã®è©³è§£ç·¨ã«è§£èª¬ã‚’ã—ã¦ã„る。 TiDBã¯MySQLã¨äº’æ›æ€§ã‚’æŒã¤ã ã‘ã§ãªãã€ã€ä»Šå›žã®è«–æ–‡ã«ç¤ºã•ã‚Œã¦ã„るよã†ã«ã€OLAP用途ã®æ©Ÿèƒ½å¼·åŒ–ã‚’è¡Œã£ã¦ãŠã‚Šã€ã“れもã¾ãŸMySQLã®å¼±ç‚¹ã‚’補強ã™ã‚‹è‰¯ã„æ–¹å‘性ã¨è¨€ãˆã‚‹ã€‚ã“ã®è¾ºã‚Šã®NewSQLã®æ©Ÿèƒ½å¼·åŒ–ã«ã¤
![[論文紹介] TiDB:a Raft-based HTAP database](https://cdn-ak-scissors.b.st-hatena.com/image/square/ad5d7368b8594081fe926e56f2548a5f3b50af26/height=288;version=1;width=512/https%3A%2F%2Fres.cloudinary.com%2Fzenn%2Fimage%2Fupload%2Fs--LwEg5Sy_--%2Fc_fit%252Cg_north_west%252Cl_text%3Anotosansjp-medium.otf_55%3A%25255B%2525E8%2525AB%252596%2525E6%252596%252587%2525E7%2525B4%2525B9%2525E4%2525BB%25258B%25255D%252520TiDB%25253Aa%252520Raft-based%252520HTAP%252520database%252Cw_1010%252Cx_90%252Cy_100%2Fg_south_west%252Cl_text%3Anotosansjp-medium.otf_37%3A%2525E3%252581%252593%2525E3%252581%2525B0%252Cx_203%252Cy_121%2Fg_south_west%252Ch_90%252Cl_fetch%3AaHR0cHM6Ly9zdG9yYWdlLmdvb2dsZWFwaXMuY29tL3plbm4tdXNlci11cGxvYWQvYXZhdGFyLzVmM2YwZjM3ZjguanBlZw%3D%3D%252Cr_max%252Cw_90%252Cx_87%252Cy_95%2Fv1627283836%2Fdefault%2Fog-base-w1200-v2.png)
How COMMIT Works in CockroachDB
CockroachDB ã¯ã©ã®ãらã„「ã—ã¶ã¨ã„ã€ã®ã‹ï¼Ÿ / How tough is CockroachDB?
Cloud Spanner ã®ãƒãƒƒã‚¯ã«ã¤ã„ã¦
ã“ã®è¨˜äº‹ã§ã¯ Cloud Spanner ã®ä¸¦è¡Œæ€§åˆ¶å¾¡ãŒä½•ã§ã‚ã‚‹ã®ã‹ã€çµæžœã¨ã—ã¦ä½•ã‚’実ç¾ã—ã¦ã„ã‚‹ã®ã‹ã‚’見ã¦ã‹ã‚‰å…·ä½“çš„ãªãƒãƒƒã‚¯ã®å®Ÿéš›ã®æŒ™å‹•ã«ã¤ã„ã¦è¿½ã£ã¦ã„ã。 ãªãŠåˆ†æ•£ã‚·ã‚¹ãƒ†ãƒ ã¨ã—ã¦ã®è©±ã¯ã‚ã¾ã‚Šãªã„ã®ã§æœŸå¾…ã—ãªã„よã†ã«ã€‚ ã“ã®è¨˜äº‹ã§ã¯å®Ÿéš›ã®æŒ™å‹•ã‚’確èªã—ãªãŒã‚‰æ›¸ã„ã¦ã„ã‚‹ã¤ã‚‚ã‚Šã ãŒã€ 2021å¹´3月ç¾åœ¨ã®æŒ™å‹•ãŒã‚µã‚¤ãƒ¬ãƒ³ãƒˆã«å¤‰ã‚ã‚‹ã“ã¨ã‚‚ã‚ã‚‹ã“ã¨ã«ã¯æ³¨æ„ã—ã¦ã»ã—ã„。 TL;DR Cloud Spanner ã¯ãƒãƒƒã‚¯ãƒ•ãƒªãƒ¼ãª Read-Only Transction ã¨ã€å …実ã«ãƒãƒƒã‚¯ã‚’行ㆠRead-Write Transaction ã®2ã¤ã®ã‚¢ã‚¯ã‚»ã‚¹ãƒ‘スをæŒã¡ã€ ROMV ã¨å‘¼ã°ã‚Œã‚‹æ–¹å¼ã«æœ€ã‚‚è¿‘ã„。 ãã®ä»–技術ã¨ã®çµ„ã¿åˆã‚ã›ã®çµæžœã¨ã—ã¦åˆ†æ•£ã‚·ã‚¹ãƒ†ãƒ ã§ã‚ã‚ŠãªãŒã‚‰ Serializability 㨠Linearizability ã‚’å…¼ãæƒãˆãŸç†è«–上最も強ã„一貫性を実ç¾ã—ã¦ãŠã‚Šã€ Google ã¯ãれを
Perceus: Garbage Free Reference Counting with ReuseMicrosoft Technical Report, MSR-TR-2020-42, Nov 29, 2020, v2.
Perceus: Garbage Free Reference Counting with Reuse Microsoft Technical Report, MSR-TR-2020-42, Nov 29, 2020, v2. Alex Reinking∗ Microsoft Research, USA [email protected] Ningning Xie∗ University of Hong Kong [email protected] Leonardo de Moura Microsoft Research, USA leonardo@microsoft.com Daan Leijen Microsoft Research, USA daan@microsoft.com Abstract We introduce Perceus, an algorithm fo
[論文紹介] Snowflake - NSDI '20 -
Building An Elastic Query Engine on Disaggregated Storage çªç„¶å§‹ã¾ã£ãŸã€è«–文紹介シリーズã§ã‚る。 データベースã«é–¢é€£ã™ã‚‹ç›®ã«ä»˜ã„ãŸè«–文をã–ã£ãã‚Šã¨è§£èª¬ã—ã¦ã„ã。個人的ãªèˆˆå‘³ã¯åˆ†æ•£DBã€ãƒˆãƒ©ãƒ³ã‚¶ã‚¯ã‚·ãƒ§ãƒ³ã€ã‚¹ãƒˆãƒ¬ãƒ¼ã‚¸ã‚¨ãƒ³ã‚¸ãƒ³ç‰ãªã®ã§ã€ãã®è¾ºã‚Šã«åã‚‹ã“ã¨ã¯ã”å®¹èµ¦é ‚ããŸã„。 åˆå›žã¯ã€ŒBuilding An Elastic Query Engine on Disaggregated Storageã€ï¼ˆæ—¥æœ¬èªžè¨³ï¼šåˆ†æ•£ã‚¹ãƒˆãƒ¬ãƒ¼ã‚¸ä¸Šã§ã®å¼¾åŠ›æ€§ã®é«˜ã„クエリエンジンã®æ§‹ç¯‰ï¼‰ã€Snowflakeã®ã‚¢ãƒ¼ã‚テクãƒãƒ£ã‚’解説ã—ãŸè«–文をèªã‚“ã§ã„ã。 ※早速ãŠè©«ã³ã¨ãªã‚‹ãŒã€Zennã®ã‚¿ã‚¤ãƒˆãƒ«ã§æ–‡å—数制é™ãŒã‚ã‚Šã€æ£å¼ãªè«–æ–‡åを当記事ã«å† ã™ã‚‹ã“ã¨ãŒå‡ºæ¥ãªã„。誤解を招ãタイトルã ã£ãŸã‚‰ç”³ã—訳ãªã„。 ã¨æ€ã£ãŸã‚‰ ã“ã¡ã‚‰ã«è«–æ–‡ã®ç¿»è¨³ãŒç™ºè¡¨ã•ã‚Œã¦ã„ãŸã€‚英語ã¯è‹¦æ‰‹ã ãŒå…¨æ–‡èªã‚“ã§ç†
![[論文紹介] Snowflake - NSDI '20 -](https://cdn-ak-scissors.b.st-hatena.com/image/square/05fa2063c17e02edf7f1a26b4992a2d1fea0db31/height=288;version=1;width=512/https%3A%2F%2Fres.cloudinary.com%2Fzenn%2Fimage%2Fupload%2Fs--PmjBGSRv--%2Fc_fit%252Cg_north_west%252Cl_text%3Anotosansjp-medium.otf_55%3A%25255B%2525E8%2525AB%252596%2525E6%252596%252587%2525E7%2525B4%2525B9%2525E4%2525BB%25258B%25255D%252520Snowflake%252520-%252520NSDI%252520%25252720%252520-%252Cw_1010%252Cx_90%252Cy_100%2Fg_south_west%252Cl_text%3Anotosansjp-medium.otf_37%3A%2525E3%252581%252593%2525E3%252581%2525B0%252Cx_203%252Cy_121%2Fg_south_west%252Ch_90%252Cl_fetch%3AaHR0cHM6Ly9zdG9yYWdlLmdvb2dsZWFwaXMuY29tL3plbm4tdXNlci11cGxvYWQvYXZhdGFyLzVmM2YwZjM3ZjguanBlZw%3D%3D%252Cr_max%252Cw_90%252Cx_87%252Cy_95%2Fv1627283836%2Fdefault%2Fog-base-w1200-v2.png)
Microsoftã®ãƒ‡ãƒ¼ã‚¿ãƒ•ãƒãƒ¼ãƒ—ãƒã‚»ãƒƒã‚µ EDGEã®è«–文をèªã‚€ - FPGA開発日記
Twitterã§çŸ¥ã£ãŸã€Microsoftã®FPGAå‘ã‘ã®ã‚¤ãƒ³ã‚ªãƒ¼ãƒ€ã‚¹ã‚«ãƒ©ãƒ—ãƒã‚»ãƒƒã‚µã®è«–æ–‡ãŒå‡ºã¦ã„る。ターゲットã¨ã—ã¦ã¯ãƒ‡ãƒ¼ã‚¿ãƒ•ãƒãƒ¼å‡¦ç†ã 。 プãƒã‚»ãƒƒã‚µã®åå‰ã¨ã—ã¦ã¯EDGEã¨ã„ã†ã€‚ブラウザã®Edgeã§ã¯ãªãã¦ã€Explicit Data Graph Executionã®ç•¥ç§°ã§ã‚る。 é¢ç™½ãã†ãªã®ã§ã€èªã‚“ã§ã¿ã‚‹ã“ã¨ã«ã—ãŸã€‚ Now Microsoft ports Windows 10, Linux to homegrown CPU design • The Register Towards an Area-Efficient Implementation of a High ILP EDGE Soft Processor https://arxiv.org/pdf/1803.06617.pdf EDGEã®è€ƒãˆæ–¹ EDGEã®è€ƒãˆæ–¹ã¯ã€è¤‡é›‘ãªã‚¢ã‚¦ãƒˆã‚ªãƒ–オーダã®ãƒ—ãƒã‚»ãƒƒã‚µã‚’è¨è¨ˆã™ã‚‹ã®ã§ã¯ãªãã€ãª
Spectre & Meltdown を防ãマイクãƒã‚¢ãƒ¼ã‚テクãƒãƒ£SafeSpecã®è«–文をèªã‚€ - FPGA開発日記
Hisa Andoæ°ã®ãƒ–ãƒã‚°ã§çŸ¥ã£ãŸã®ã ãŒã€Spectre & Meltdownを防ãマイクãƒã‚¢ãƒ¼ã‚テクãƒãƒ£ã¨ã—ã¦SafeSpecã¨ã„ã†æŠ€è¡“ãŒç™ºè¡¨ã•ã‚ŒãŸã®ã§ã€ã“れをèªã‚“ã§ã¿ã‚‹ã“ã¨ã«ã—ãŸã€‚ ２．Spectreã‚„Meltdownを防ãSafeSpecマイクãƒã‚¢ãƒ¼ã‚テクãƒãƒ£ 20180616 Boffins offer to make speculative execution great again with Spectre-Meltdown CPU fix www.theregister.co.uk Meltdown/Spectreã®å¯¾ç– SafeSpec keisanki.at.webry.info è«–æ–‡ 「SafeSpec: Banishing the Spectre of a Meltdown with Leakage-Free Speculation〠https://arxiv.o
深層å¦ç¿’ã®ãƒ¢ãƒ‡ãƒ«åœ§ç¸®ãƒ»é«˜é€ŸåŒ–ã«é–¢ã™ã‚‹è«–æ–‡80本ノック - ãã†ã ã
ã“ã‚“ã«ã¡ã¯.無事ã«3年生ã«é€²ç´šã§ãã¦ãƒ›ãƒƒã¨ã—ã¦ã„ã¾ã™. 昨年度ã‹ã‚‰æ·±å±¤å¦ç¿’ã®ãƒ¢ãƒ‡ãƒ«æ”¹å¤‰ã«ã‚ˆã‚‹é«˜é€ŸåŒ–手法ã«ã¤ã„ã¦èª¿ã¹ã¦ã„ã¾ã—ãŸ.ãã‚“ãªã‚ã‘ã§æœ¬ç¨¿ã§ã¯æ·±å±¤å¦ç¿’モデルã®ãƒ‘ラメータ容é‡ã®å‰Šæ¸›ã‚„高速化を目的ã¨ã—ãŸæ‰‹æ³•ã«é–¢ã™ã‚‹è«–文を80本紹介ã—ã¾ã™. 一部, ãƒã‚¤ãƒ‘ーパラメータ探索(Neural Architecture Searchç‰)ã®è«–文を「高速ãªãƒ¢ãƒ‡ãƒ«ã®æŽ¢ç´¢ã«ä½¿ãˆãªã„ã‹ãªã€œã€ã¨ã„ã†ãã‚‚ã¡ã§å«ã‚ã¾ã—ãŸ. ã¾ãŸ, 蒸留ã«é–¢ã™ã‚‹è«–æ–‡ã¯å¿…ãšã—も高速化を狙ã£ãŸã‚‚ã®ã ã‘ã§ã¯ãªã, 蒸留ãã®ã‚‚ã®ã®æ´¾ç”Ÿè«–æ–‡ã«ã¤ã„ã¦ã‚‚å«ã‚ã¦ã„ã¾ã™. å…¨ã¦ã®è«–æ–‡ã«ã¤ã„ã¦è©³è§£ã§ã¯ãªã雑ãªæ¦‚è¦ã®ã¿è¨˜è¼‰ã—ã¦ã„ã¾ã™. サーベイã™ã‚‹éš›ã®ã‚¬ã‚¤ãƒ‰ã«ã§ã‚‚使ã£ã¦ã„ãŸã ã‘ã‚Œã°å¹¸ã„ã§ã™. サクッã¨èªã‚“ã ã ã‘ã®è«–文も多々ã‚ã‚Šã¾ã™ã®ã§é–“é•ã„ç‰ã”ã–ã„ã¾ã—ãŸã‚‰ã”指摘ãã ã•ã„. ã§ã¯å‚ã‚Šã¾ã—ょã†. 高速ãªãƒ¢ãƒ‡ãƒ«ã‚¢ãƒ¼ã‚テクãƒãƒ£ より良ã„速度ã¨ç²¾åº¦ã®ãƒˆãƒ¬ãƒ¼ãƒ‰ã‚ªãƒ•ã‚’
世界ã§æœ€ã‚‚引用ã•ã‚ŒãŸè«–文トップ100件ã®ã†ã¡65件ã¯æœ‰æ–™åŒ–ã•ã‚Œã¦ã„ã¦ã€Œèª°ã§ã‚‚アクセスã§ãる状態ã€ã§ã¯ãªã„
by Shawn Campbell インターãƒãƒƒãƒˆã¯ã‚‚ã¨ã‚‚ã¨å¤§å¦ã®ã‚³ãƒ³ãƒ”ューターåŒå£«ã‚’ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã§æŽ¥ç¶šã—ã¦è«–文を共有ã™ã‚‹ãŸã‚ã«ä½œã‚‰ã‚ŒãŸã‚‚ã®ã§ã™ãŒã€ã“ã®ç¬¬ä¸€ç›®æ¨™ã«åã—ã¦å¼•ç”¨ä»¶æ•°ãƒˆãƒƒãƒ—100ã®è«–æ–‡ã®å¤§åŠãŒã‚ªãƒ¼ãƒ—ンアクセスã§ã¯ãªã„状態ã«ã‚ã‚‹ã“ã¨ãŒæŒ‡æ‘˜ã•ã‚Œã¦ã„ã¾ã™ã€‚ 65 out of the 100 most cited papers are paywalled. https://www.authorea.com/users/8850/articles/125400-65-out-of-the-100-most-cited-papers-are-paywalled 指摘を行ã£ãŸã®ã¯ã€ç ”究者å‘ã‘ã®åŸ·ç†ãƒ„ール「Authoreaã€ã®å…±åŒå‰µæ¥è€…ã§ã‚るアルベルト・ãºãºæ°ã¨ã€ä¸»ä»»ç ”究者ã®ã‚¸ãƒ§ã‚·ãƒ¥ãƒ»ãƒ‹ã‚³ãƒ«ã‚½ãƒ³æ°ã€‚ è«–æ–‡ã«ã‚ˆã£ã¦ä¾¡æ ¼ã¯ç•°ãªã‚‹ã‚‚ã®ã®ã€åŠ é‡å¹³å‡ã®çµæžœã¯32.33ドル(ç´„3620円)ã§ã—ãŸã€‚ä»®ã«ã€æœ‰
Cicada：Dependably Fast Multi-Core In-Memory Transactions - 急ãŒã°å›žã‚Œã€é¸ã¶ãªã‚‰è¿‘é“
Cicada: Dependably Fast Multi-Core In-Memory Transactions https://www.cs.cmu.edu/~hl/papers/cicada-sigmod2017.pdf SIGMOD2017ã§ç™ºè¡¨ã•ã‚Œã¦ã„る。ç¾çŠ¶ã®åˆ†æ•£OLTPã®ã‚¢ãƒ¼ã‚テクãƒãƒ£ã‚’ã†ã¾ãã¾ã¨ã‚ã¦ã€æ¬ 点をã†ã¾ãã‚«ãƒãƒ¼ã‚¢ãƒƒãƒ—ã—ã€è¨€ã£ã¦è¦‹ã‚Œã°æ¬¡ä¸–代MVCCã®ä¸€ã¤ã®å½¢ã‚’æ示ã—ã¦ã„る。ãã®ä¸Šã§ã€ç¾åœ¨ä¸–界最高ã®ãƒ‘フォーマンスをå©ã出ã—ã¦ã„る。ç¾æ™‚点ã§ä¸–界最速DB（ãŸã ã—自称）。 ç¾çŠ¶ã®åˆ†æ•£OLTPã¯å¤§ããªæµã‚Œã¯ã€SILO/Foedus/MOCC/ç‰ã®OCCç³»ã€ã™ãªã‚ã¡2PLをベースã«ã—ãŸå®Ÿè£…ã§ç†è«–上ã¯monoversionã§ã®serializableã®å®Ÿç¾ã‚’è¡Œã£ã¦ã„ã‚‹æ–¹å¼ã¨ã€Hekaton/HyPer/Bohm/ERMIAã¨ã„ã£ãŸMVCCç³»ã€ã™ãªã‚ã¡MVTOã®æ´¾ç”Ÿã‚’ベースã«ã—ã¦mu
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
Amazon Redshift Re-invented
File System Semantics Requirements of HPC Applications
https://www.ipsj-kyushu.jp/page/ronbun/hinokuni/1007/B1/B1-4.pdf
SQL- and Operator-centric Data Analytics in Relational Main-Memory Databases
The Battle of the Schedulers: FreeBSD ULE vs. Linux CFS
Apache Calcite: A Foundational Framework for Optimized Query Processing Over Heterogeneous Data Sources (PDF)
Cluster-Level Storage @ Google
Mison: A Fast JSON Parser for Data Analytics
Azure Data Lake Store: a hyperscale distributed file service for big data analytics | the morning paper
https://doi.org/10.1145/3035918.3056100
{{#tags}}- {{label}}
{{/tags}}