Energy Measurements for Cassandra Cloud Storage System: Exploring and improving Energy-Consistency Tradeoff - DUMAS - Dépôt Universitaire de Mémoires Après Soutenance
Master Thesis Year : 2013

Energy Measurements for Cassandra Cloud Storage System: Exploring and improving Energy-Consistency Tradeoff

Abstract

Cassandra is an open-source cloud storage system that offers three types of consistency per-operation including weak, eventual and strong consistency. It is being used by many data-center applications (e.g., Facebook, AppScale). Most existing research efforts have been dedicated on exploring and improving consistency-performance, consistencylatency and consistency-monetary cost tradeoffs. In contrast, little work is focusing on energy-consistency tradeoff. As power bills have become a substantial part of the monetary cost for operating a data-center, this report aims at providing more clear understanding of the interplay between consistency and power consumption. Accordingly, a serious of experiments have been conducted to explore the implication of different factors including the selected consistency level and the access pattern and the load of the running applications on the energy-consumption. Our experiments have revealed a noticeable energy consumption variation according to the selected consistency and the access patten and load exhibits by the application. Moreover, we observe that this variation is contributed to by the obtained throughput of the applications and also by the uneven distribution of the loads amongst different nodes in the Cassandra system. Accordingly, we have proposed three data-layout policies and we evaluate the energy-consumption of these policies in Cassandra through experiments. Interestingly, we have achieved 24% energy saving when assigning 50% of the nodes to the warm pool for the applications with moderate ratio of reads and writes when applying eventual (Quorum) consistency.
Fichier principal
Vignette du fichier
YueLi.pdf (3.09 Mo) Télécharger le fichier
Loading...

Dates and versions

dumas-00854995 , version 1 (28-08-2013)

Identifiers

  • HAL Id : dumas-00854995 , version 1

Cite

Yue Li. Energy Measurements for Cassandra Cloud Storage System: Exploring and improving Energy-Consistency Tradeoff. Distributed, Parallel, and Cluster Computing [cs.DC]. 2013. ⟨dumas-00854995⟩
372 View
287 Download

Share

More