Improving Throughput in Cloud Storage System
Journal Title: International Journal on Computer Science and Engineering - Year 2012, Vol 4, Issue 6
Abstract
Because the cloud serves many workloads concurrently, its disk access pattern is highly random and heterogeneous. In addition, because various virtual machines access to files respectively, meta data utilization and small write requests are increased. In order to build a system for these patters, we should analyze the modern techniques used at cloud system. First, we show that a SATA controller has enough processing capability to serve six disks without performance degradation. Motivated by this experiment, we compare three disk configurations by executing benchmark applications. From the result, we observe that accessing independent disks directly provides better performance than RAID-0 and RAID-5. Lastly, we performed experiment about comparing frequently used protocols (NFS, iSCSI)[8] and got the result that iSCSI[10] showed better performance than NFS[11]
Authors and Affiliations
Chanho Choi , Shin-gyu Kim , Hyeonsang Eom , Heon Y. Yeom
Keywords
Related Articles
- EP ID EP86811
- DOI -
- Views 137
- Downloads 0
Improving the Performance of K-Means Clustering For High Dimensional Data Set
Clustering high dimensional data is the cluster analysis of data with anywhere from a few dozen to many thousands of dimensions. Multiple dimensions are hard to think in, impossible to visualize, and, due to the exponent...
FUZZY LOGIC BASED SOFTWARE PROCESS IMPROVIZATION FRAMEWORK FOR INDIAN SMALL SCALE SOFTWARE ORGANIZATIONS
In this paper, the authors elaborate the results obtained after analyzing and assessing the software process activities in five small to medium sized Indian software companies. This work demonstrates a cost effective fra...
AI based Digital Companding Scheme for OFDM system using custom constellation Mapping and selection
Data rate is important in telecommunication because it is irectly proportional to the cost of transmitting the signal. Saving bits is the same as saving money . In this paper we propose new digital companding scheme for...
Learning Approaches toward Title Word Selection on Indic Script
Title is a compact representation of a document which distill the important information from the document. In this paper we studied the selection words as title words by using different learning approaches namely nearest...
Optimizing Hidden Markov Model for Failure Prediction– Comparison of Gaine’s optimization and Minimum message length Estimator
Computer systems are prone to failures. Failures are caused by faults that occur in a system. As faults are unknown and cannot be measured, they produce error messages on their detection. The approach presented here is t...