@article{oai:ipsj.ixsq.nii.ac.jp:00162233, author = {永田, 正樹 and 阿部, 祐輔 and 金原, 一聖 and 福井, 美彩都 and 峰野, 博史 and Masaki, Nagata and Yusuke, Abe and Issei, Kinpara and Misato, Fukui and Hiroshi, Mineno}, issue = {1}, journal = {情報処理学会論文誌コンシューマ・デバイス&システム(CDS)}, month = {May}, note = {本論文では,WEBシステムを構成するサーバ群を世界規模で冗長化し,アクセス数を事前予測して適切なサーバ数で負荷分散を行う広域冗長型安否システムを提案する.既存の安否システムには,災害時の確実な稼働とアクセス状況に応じた適切なサーバ数算出が困難という課題がある.そこで提案システムでは,日本,シンガポール,カリフォルニア,など多地点および地理的に離れた広域拠点にシステムを分散配置し可用性を向上させる広域冗長化機構と,災害規模に応じたアクセス数を事前予測し適切なサーバ数で負荷分散を行うアクセス予測に基づくオートスケーリング機構を実装した.過去の災害データを基にしたシミュレーション評価の結果,多地点での広域冗長化稼働とアクセス予測に基づくオートスケーリング機構を用いて年間費用約32%削減の見込みを得ることでき,提案システムの有効性を確認した., In this paper, we propose global redundant web safety information system which implemented global redundant on the world area and load balancing by the access prediction model. Conventional web safety information system has a problems, which are certain operation at the disaster and adjustment of the number of servers in conformity with access. So we have developed global redundant and auto scaling based on access prediction model to settle a problems. Global redundant arranges a system separately in the distant area and multiple locations (e.g., Japan, California and Singapore). Auto scaling based on access prediction model predicts the access amount according to the disaster scale before access concentration. And load balancing executes. A result of the simulation experiment expected a cost reduction about 32% by global redundant and auto scaling based on access prediction model, and the validity of the proposal system was confirmed.}, pages = {94--105}, title = {アクセス予測に基づいた広域冗長型安否システムの提案と基礎評価}, volume = {6}, year = {2016} }