Languages

You are here

Cloud

Group: Cloud

Research introduction

In Application Performance Management (APM), the most common problem encountered by the administrators of many network service providers is how to sustain performance targets for the applications. Cloud computing offers the possibility to provide the resources on demand as Utility Computing. In cloud group,   we work on three major projects to solve the individual problems.

 

Firstly, we work on a project which is developing a mechanism that dynamically allocates resources to application on demand. The reason why we need such a mechanism is that through a better arrangement (allocation) of resources (includes CPU, memory, disk, network), applications can benefit a better performance and even save more money by turning off unused resources.  In this project, we introduce a four-steps procedure as below to achieve our goal.  From (A), we monitor the workload and measure performance metrics of the web application. We can check the measures against target performance and record the external demand for prediction model. Next in (B), a prediction algorithm uses the online measurements from the first phase to estimate the external demand in near future. In the next phase (C), we take an analytic performance model which can use the estimated external demand to determine the resource requirement for specific application. Last phase (D), the resource requirement is translated to the system controllable resources configuration, and then the system automatically reconfigures resource allocation for the application to meet its SLA performance goals with time-varying workloads.  

 

 

 

 

 

 

 

 

 

 

 

 

Secondly, in cloud computing, in order to maximize the utilization of resources, we use the virtualization technology. Though we can benefit a better utilization of resources, there are evitable overheads incurs from virtualization. So we want to develop a Workload-Performance-Resource Model for cloud service provider (CSP). By using the model, CSP could rapidly make an accurate resource allocation for a new incoming application to cloud environment by its target performance and known workload.

 

 

Thirdly, we aim to exploit the advantage of resource provision in the cloud. To help applications with deadline (AppD) get resources needed. In this project, we expect to  provide 2 main capabilities: 1) Providing resource needed instantly, 2) Efficiently allocating resource to application with deadline. In order to achieve objectives above, we design mechanisms that monitor application's progress and scaling up/down cloud resources dynamically. With AppD, users could complete their work in time conveniently

 

在應用效能管理(Application Performance Management),許多網絡服務提供商的管理員所遇到的最常見的問題是有效的利用資源並保持應用程式的效能目標。雲端運算提供了有效率的解決方案。在雲端組中,我們分為三大方向的研究用以解決問題。

 

首先,我們正在開發一種機制,動態地依照需求將資源分配給應用程序。我們之所以需要這樣的機制是因為通過一個更好的安排(分配)資源(包括CPU、記憶體、儲存空間、網路),應用程序可以獲得更好的性能,甚至關閉未使用的資源,節省更多的錢。在這個項目中,我們引入了一個四步驟的流程以實現我們的目標。 在(A)中,我們監測Web應用程式的workload和衡量效能的指標。我們可以將在這階段中蒐集到的目標效能指標和workload記錄,當作input用在外部需求預測模型(B)中。接著,在(B)中藉由從第一階段蒐集的資訊,online的用來估計在不久將來的外部需求。 (C)在下一階段,我們的分析性能模式中,可以使用估計的外部需求來確定特定應用的資源需求。最後一個階段(D),資源要求被翻譯系統的可控制的資源配置,然後系統會自動重新配置資源分配的應用程序,以滿足其SLA隨時間變化的工作負載的性能目標。

 

 

接下來,在雲端運算中,為了能夠達到資源利用的最大化,虛擬化是一個非常有用的技術,但是虛擬化的技術無可避免的會帶來一些額外overhead,因此,為了能將應用程式從實體環境中搬遷到虛擬機器上,且在搬遷之後盡可能的不影響應用程式的效能及使用者訂下的SLA,本研究根據在原本的實體環境及虛擬機器上觀測到的效能參數,為各種不同類型的應用程式建立有關負載、效能、資源的「概要分析」(profiling)。憑藉著此profile,能夠評估出在虛擬環境裡需要配置多少資源才能夠讓從實體環境中搬遷的應用程式達到同樣的效能或是SLA,此profiling的產出又可稱為Workload-Performance-Resource Model(WPR Model)。

 

最後,我們專注在需要大量運算資源且附有執行期限的應用程式(application with deadline,  AppD)。為了發揮雲端在資源供應上的優勢,讓有執行期限的計算工作能取得所需要的大量資源,我們提供兩個重要的功能:1) 能夠快速地產生需要的資源,2) 能有效地配置資源給有執行期限的工作。為達成以上目標我們設計一些機制來動態的監控工作進度和規模化資源配置。期望AppD能幫助使用者輕鬆又方便地在時限內完成工作

 

 

 

PublicationTools sharing

目前沒有@@”

 

Current members

http://sphotos-g.ak.fbcdn.net/hphotos-ak-ash3/77129_159903994052023_5730916_n.jpg

姓名 : 陳昱帆  Yu-Fan Chen

Email: r00725026 (at) ntu.edu.tw

Information Management, National Central University, Taiwan

Tel: +886-988-663-502

 

 

 

 

 

 

http://140.112.161.227/odsainfo/getpic.aspx?reg_no=30DE7E72594B6E3B7215806A6DC38E49姓名:杜承恩 Cheng-En, Du

Email:briandododo (at) gmail.com

Information Management, National Central University, Taiwan

Tel:+886-970-791-796

 

 

 

 

 

 

 

 

Mac:Users:weiru:Desktop:G1:戴瑋如.jpg                           

姓名: 戴瑋如  Wei-Ru Dai

Email: r01725012 (at) ntu.edu.tw

Information Management, National Taiwan University, Taiwan

Tel: +886-2-2788-3799#1614

 

 

 

 

 

 

姓名: 吳張民  Chang-Ming Wu

Email: r01725027 (at) ntu.edu.tw

Information Management, National Taiwan University, Taiwan

Tel: +886-2-2788-3799#1614

 

 

 

Undefined
Copyright 2013 © Advanced Networking Technologies and Services Group, ALL RIGHT RESERVED.