講演名 2018-07-30
Analysis of commercial cloud workload and study on how to apply cache methods
大江 和一(富士通研), 荻原 一隆(富士通研), 本田 岳史(FST),
PDFダウンロードページ PDFダウンロードページへ
抄録(和)
抄録(英) We analyzed some storage workloads of the FUJITSU K5 cloud service, which was built using theOpenStack platform, to clarify how to handle these workloads effectively. We retrieved these storage workloads andanalyzed them from the viwe point of both temporal and spatial access locality. For temporal access locality, weclassified these workloads as two input/output (IO) access patterns. One was that the number of IO accesses washigh only for a specific time and the other was that the number of IO accesses was almost stable. For both temporaland spatial access locality, we found that almost all workloads included IO concentrations. Such concentrationsare aggregations of IO accesses and appear in narrow regions of a storage volume and continue for periods of upto an hour. Therefore, we studied how to apply cache methods by using a cache simulator. The cache hit ratiosvaried with workload, but almost half of the workloads had low cache hit ratios because they included few page-levelregularities. Automated tiered storage with fast memory and slow flash storage (ATSMF), which was we proposedin our previous study, can handle IO concentration effectively because its migration algorithm found the entire areaof IO concentration and migrated it from slow storage to fast storage if the migration improves user’s performance. We predict that the cache hit ratios of ATSMF will be much higher than the half workloads of the Least RecentlyUsed and Adaptive Replacement Cache algorithms. Therefore, we should develop a technique that combines ATSMFand traditional caching. From the results of temporal access locality, we can a share fast storage area among severalworkloads when these workloads have different times when the number of IO accesses is high. Moreover, we shouldpreferentially allocate a fast storage area to IO concentrated workloads, which accounts for only 1.2% of the totalworkloads, because these workloads include more than half of all IO accesses.
キーワード(和)
キーワード(英) Cloud workloadanalysiscachingLRUARCIO concentrationATSMF
資料番号 CPSY2018-14
発行日 2018-07-23 (CPSY)

研究会情報
研究会 CPSY / DC / IPSJ-ARC
開催期間 2018/7/30(から3日開催)
開催地(和) 熊本市国際交流会館
開催地(英) Kumamoto City International Center
テーマ(和) 並列/分散/協調システムとディペンダブルコンピューティングおよび一般
テーマ(英) Parallel, Distributed and Cooperative Processing Systems and Dependable Computing
委員長氏名(和) 中野 浩嗣(広島大) / 福本 聡(首都大東京) / 五島 正裕(NII)
委員長氏名(英) Koji Nakano(Hiroshima Univ.) / Satoshi Fukumoto(Tokyo Metropolitan Univ.) / Masahiro Goshima(NII)
副委員長氏名(和) 入江 英嗣(東大) / 三吉 貴史(富士通研) / 高橋 寛(愛媛大)
副委員長氏名(英) Hidetsugu Irie(Univ. of Tokyo) / Takashi Miyoshi(Fujitsu) / Hiroshi Takahashi(Ehime Univ.)
幹事氏名(和) 大川 猛(宇都宮大) / 高前田 伸也(北大) / 金子 晴彦(東工大) / 新井 雅之(日大) / 小野 貴継(九大) / 近藤 正章(東大) / 長谷川 揚平(東芝) / 塩谷 亮太(名大)
幹事氏名(英) Takeshi Ohkawa(Utsunomiya Univ.) / Shinya Takameda(Hokkaido Univ.) / Haruhiko Kaneko(Tokyo Inst. of Tech.) / Masayuki Arai(Nihon Univ.) / Takatsugu Ono(Kyushu Univ.) / Masaaki Kondo(Univ. of Tokyo) / Yohei Hasegawa(Toshiba) / Ryota Shioya(Nagoya Univ.)
幹事補佐氏名(和) 伊藤 靖朗(広島大) / 津邑 公暁(名工大)
幹事補佐氏名(英) Yasuaki Ito(Hiroshima Univ.) / Tomoaki Tsumura(Nagoya Inst. of Tech.)

講演論文情報詳細
申込み研究会 Technical Committee on Computer Systems / Technical Committee on Dependable Computing / Special Interest Group on System Architecture
本文の言語 ENG
タイトル(和)
サブタイトル(和)
タイトル(英) Analysis of commercial cloud workload and study on how to apply cache methods
サブタイトル(和)
キーワード(1)(和/英) / Cloud workloadanalysiscachingLRUARCIO concentrationATSMF
第 1 著者 氏名(和/英) 大江 和一 / Kazuichi Oe
第 1 著者 所属(和/英) 富士通研究所(略称:富士通研)
Fujitsu Laboratories LTD.(略称:FLABO)
第 2 著者 氏名(和/英) 荻原 一隆 / Kazutaka Ogihara
第 2 著者 所属(和/英) 富士通研究所(略称:富士通研)
Fujitsu Laboratories LTD.(略称:FLABO)
第 3 著者 氏名(和/英) 本田 岳史 / Takeo Honda
第 3 著者 所属(和/英) 富士通ソフトウェアテクノロジーズ(略称:FST)
FST(略称:FST)
発表年月日 2018-07-30
資料番号 CPSY2018-14
巻番号(vol) vol.118
号番号(no) CPSY-165
ページ範囲 pp.7-12(CPSY),
ページ数 6
発行日 2018-07-23 (CPSY)