Comparing serial and parallel compressive sensing for internet traffic matrix
Irawati I.D.a, Suksmono A.B.a, Edward I.J.M.a
a School of Electrical and Informatics, Institut Teknologi Bandung, Indonesia
Abstract
© 2018 Newswood Limited. All rights reserved.Compressive Sensing (CS) is a new method capable of efficiently reconstructing signals by using sparse sample. However, CS algorithms require processing time very extensive especially since the amount of data is very large. In this paper, we evaluated the effect of using double CS processes either serial CS (SCS) and parallel CS (PCS) on Internet traffic matrix. We also compared two reconstruction algorithms, which are Orthogonal Matching Pursuit and Iteratively Reweighted Least Square (IRLS). SCS produces poor accuracy with longer processing time, while PCS produce accuracy similar to CS scheme with shorter processing time. We also examine the effect of subparallel on the performance results. The results show that the greater number of subparallel accelerate the processing time for IRLS, contrary to OMP, where more subparallel, decreasing accuracy.
Author keywords
Accuracy,Compressive sensing,Parallel,Processing time,Serial,Subparallel
Indexed keywords
Accuracy,Compressive sensing,Parallel,Processing time,Serial,Subparallel
Funding details
Manuscript received December 08, 2017; revised January 29, 2018. This work was supported in part by the Telkom Foundation and LPPM ITB Indrarini Dyah Irawati. Author is with School of Electrical and Informatics, Institut Teknologi Bandung, Indonesia (phone: 62-8122152542; fax: 62-22-7507712; e-mail: indrarini@ telkomuniversity.ac.id).