透過您的圖書館登入
IP:3.16.135.254
  • 學位論文

考量依時間變化負載於資料中心內社群服務節能資料放置方法

Energy-efficient Data Placement with Time-variant Workloads Consideration for Social-aware Service in Data Centers

指導教授 : 周承復

摘要


近年來社群網路服務的快速興起,大部分民眾已經習慣於使用這些社群網路服務來與親朋好友互動,社群網路服務提供者也相對應的使用越來越多的機器以及資源來保證能滿足大量使用者服務需求,但這些社群服務的使用量並不會永遠處在高檔,其實在大部份的時間中整個提供社群網路服務的系統所需要的資源只有全部資源中的一部份而已,這相對而言持續占用全部的資源也造成了資源上的浪費以及金錢上的損失,因此我們可以去在適當的時機去減少一些資源的消耗,已達到更高的能源運用效率。 在本篇論文中,我們從新浪微博的資料當中發現並驗證了這個系統服務負載的變化趨勢以及關係,我們因而去提出了一套考慮系統負載變化量去決定留用多少數量的機器來提供服務,以及考慮時間而造成使用者使用狀態變化的資料放置方法,同時在資源使用量較少的情況下,我們也應該去保證整個系統也能提供一定的服務品質,讓使用者不會受太大的影響,而我們也受惠於能在適當時機去減少資源使用量的原因,我們因而能達到更好的能源使用效率。

並列摘要


In recent years, the social aware services become more and more prevalent. Lots of people interact with their friends and family through the social aware service. As a consequence, providers of the social aware services apply a lot of resources to serve their users. However, the workload of the social aware services is not always at the peak. In fact, the services only take part of the resources to work well at lots of the time. And this phenomenon will cause the waste of the resources. Hence, we can reduce the usage amount of the resources to be more energy efficient. In this paper, we observe the trend and change of the workload of the data of a social aware service, Sina Weibo. We then propose a method to decide the number of machines we will use with regard to the workload change of the system, and we partition the social graph with the considering of different usage patterns of users in different time. At the same time, we need to ensure a level of QoS of the system, while we use fewer resources. And we can also be more energy efficient because we consume fewer resource

參考文獻


1. Facebook Reports Fourth Quarter and Full Year 2013 Results. Available from: http://investor.fb.com/releasedetail.cfm?ReleaseID=821954.
3. Apache cassandra.
4. Mongodb. Available from: http://www.mongodb.org/.
5. METIS - Serial Graph Partitioning and Fill-reducing Matrix Ordering. Available from: http://glaros.dtc.umn.edu/gkhome/metis/metis/overview.
8. Josep, M.P., et al., The little engine(s) that could: scaling online social networks, in Proceedings of the ACM SIGCOMM 2010 conference

延伸閱讀