IMI Colloquium

4月 IMI Colloquium

日時：2013年4月17日（水）
16:45-17:45
場所：九州大学 マス・フォア・インダストリ研究所
大講義室１（数理・IMI図書館棟３F）

講師：山本 政行 氏（株式会社日立製作所 横浜研究所 ストレージシステム研究部 部長）
講演タイトル：ストレージシステムと研究概要 (10分)

講師：中村 隆喜 氏（東北大学 電気通信研究所）
講演タイトル：大規模災害でのデータ可用性を向上するリスクアウェア複製機能の提案 (20分)

講演要旨： 東北大学電気通信研究所は，文科省委託事業「高機能高可用性ストレージ基盤
技術の開発」を，日立製作所，日立ソリューションズ東日本と共に，平成24年度より5ヵ年
計画 で推進中である．同事業では，東日本大震災のような地域ごと壊滅するような大規模
災害においても，データが失われず継続してアクセスが可能なシステムの実現を目指してい
る．この実現手段の一つとして，我々はデータ消失リスクを考慮してデータ複製先を選択す
るリスクアウェア複製機能を提案中である．本発表では，複製先選択に用いる数 理モデル
と，線形計画法を活用した複製先決定アルゴリズムを紹介する．

Speaker：Masayuki Yamamoto (Manager, Data Storage Systems Research Dept, Yokohama Research Laboratory, Hitachi, Ltd.)
Title：Storage systems and the research outline

Speaker：Takaki Nakamura (Research Institure of Electrical Communication, Tohoku University)
Title：A proposal of risk-aware replication to improve data availability under wide-spread disasters
Abstract:
Research Institute of Electrical Communication (RIEC), Tohoku University is engaged in a national project "Development of Storage Platform Technology with High-Functionality and High-Availability" commissioned
by MEXT in collaboration with Hitachi, Ltd. and Hitachi Solutions East Japan, Ltd. from FY2012 to FY2017.
One of our goals in this project is to realize a system which can avoid data loss and continue to access
the data even under a wide-spread disaster like the Great East Japan Earthquake which damages wide
area. To achieve the goal, we propose risk-aware replication which selects replication target with
consideration of data loss risk. In this talk, we introduce mathematical model for selection of replication
target and decision algorithm by using linear programming.

Speaker：Hiroaki Akutsu
(Data Storage Systems Research Dept, Yokohama Research Laboratory, Hitachi, Ltd.)
Title：Technology of Data Relocation Process in Automated Storage Tiering Functions
Abstract:
The era of Big Data has arrived, and the increase of the amount of data and diverse types of data is in
progress. And the diversification of the storage media type in the storage caused by the spread of using
flash memory devices, it becomes to have difficulty in performance design and data placement to the
appropriate tier by the hand operation. This problem begins to be solved by Automated Storage Tiering
Functions. As for Automated Storage Tiering Functions, small pieces of data are automatically relocating
among the tier by access frequency. In this lecture, we introduce the technology of data relocation process
in Automated Storage Tiering Functions. In particular, we show the behavior of scheduling schemes of data
copy in relocation process by analytic models based on queuing theory.

* In this IMI Colloquium, we hold three lectures about "Application of the mathematical science
techniques to the research of storage systems".