標題: 結合雲端運算與感測網路之應用服務平台-子計畫三:SeC-Plat以感測網路為基礎之雲端資料查詢索引及探勘平台
SeC-Plat: Efficient Query Processing and Dynamic Resource Management for Sensor enabled Cloud Computing Platform
作者: 彭文志
Peng Wen-Chih
國立交通大學資訊工程學系(所)
關鍵字: 雲端資料處理;資料管理;資料查詢;索引建立
公開日期: 2013
摘要: 近年來雲端運算科技的進步,提供運算能力已成為一種服務。雲端運算平台提 供大規模儲存與資料運算分析平台,現有的雲端運算應用均限定於靜態的文件儲存、網 頁關連性分析等。我們預期未來的雲端應用將與實際的生活環境更為息息相關,如透過 感測器(Sensor)的佈建,感測器透過無線網路 (Wireless Networks) 互相連結,負責監控 並蒐集生活環境相關的資料,此一環境架構即為無線感測網路環境 (Wireless Sensor Networks)。透過無線感測網路收集資料,送達雲端運算平台做資料收集與分析,將是未 來雲端運算的主要應用。在本計劃中,感測網路著重在車載感測 (含視訊)以及使用者透 過手持裝置上之感測裝置,在感測資料的收集與分析上將是大規模 (Large-Scale) 的情 境,我們的計畫目標在於建構一個以感測網路為基礎的雲端資料查詢索引及探勘平台, 提供與使用者生活環境周遭有關之雲端應用服務。 本計畫是為期三年之總計畫『結合雲端運算與感測網路之應用服務平台』之子計畫 三 『SeC-Plat: 以感測網路為基礎的雲端資料查詢索引及探勘平台』。在此子計畫中, 我們將研發適合於感測網路之雲端運算平台。具體而言,感測資料具備了空間維度之資 訊,針對某個空間查詢範圍,運算平台當迅速回傳所有落在這區域的感測資料給雲端應 用程式。然現行列式(Colum-oriented)雲端資料庫系統尚未支援空間資料的索引結構,因 此,在第一年,我們將研發雲端運算平台之空間索引結構。在第二年,我們將研發雲端 運算平台之動態虛擬機器配置機制,以充分利用雲端運算資源於感測資料收集與探勘。 有鑑於雲端應用服務,將有大量的資料查詢被執行,因此,在第三年中,我們將研究雲 端運算平台之資料查詢最佳化機制。整體而言,我們主要的研究課題有三:(1) 針對現 行雲端資料庫多屬於列式資料庫,設計空間索引結構,提升區域查詢(Range Query)的效 率; (2)監控運算平台上每個虛擬機器的資料量與運算負荷量,設計動態虛配置擬機器的 機制,使得虛擬機器能適當配置給來自應用服務端的每個查詢,避免造成運算資源的浪 費; (3)當雲端運算平台同時間接收到大量查詢時,我們當設計查詢最佳化機制,分析查 詢間的重複性來避免相同子查詢被重複執行的情形,提升系統整體的效能。 隨著無線感測網路的進步與雲端運算近年的蓬勃發展,我們相信此計畫之執行, 將可研發出適用於感測網路為基礎之雲端運算平台,提供資料查詢索引與探勘之前瞻性 技術。
With the advance of cloud computing technology, cloud computing becomes a powerful way to provide large-scale computing and analysis. It is shown that cloud computing can efficiently store and analyze static documents and Web pages. We expect that in the future, cloud computing is able to play an important platform to collect and analyze sensor data sensed from physical worlds. In our project, we focus on sensor data from vehicles and mobile phones since the number of vehicles and mobile phones is intrinsically huge, which could demonstrate the scalability in data collection and analysis over cloud computing platforms. This project is under the integrated project 『SeC: Sensor-Enabled Cloud Service Platforms』, and aims at designing cloud data management that consists of index structures, query optimizations and mining mechanisms. Our primary goals include (1) spatial index structures; (2) dynamic virtual machine managements; (3) multiple query optimizations. More specifically, due to that sensor data usually has spatial information; range queries for sensor data will become a basic query type in sensor-enabled cloud service platforms. However, current cloud data management does not provide such spatial index structures. Thus, in the first year, we intend to develop a spatial index structure for range queries and the developed spatial index structure will be implemented on key-valued column-oriented cloud databases. In the second year, we develop a dynamic virtual machine (VM) management. Since cloud computing platforms have their own VMs to serve a variety of cloud services, our proposed management is able to efficiently balance workloads of VMs to guarantee good qualities for cloud services. Moreover, in cloud computing platforms, a huge amount of queries may be submitted. To further improve the system performance, in the third year, we propose a query optimization mechanism in cloud computing platforms. In view of the increasing attention on cloud computing, we strongly believe that this project is very timely and will deliver results of both theoretical and practical importance.
官方說明文件#: NSC100-2218-E009-016-MY3
URI: http://hdl.handle.net/11536/92537
https://www.grb.gov.tw/search/planDetail?id=2849520&docId=403339
Appears in Collections:Research Plans