A Sparse Dirty Data Process Method Based on Alternative Projection for Traffic Matrix Estimation
Feng Yu, Wei Li, Zhang Zaiyue, Jiang Lalin · 2009
Estimation of traffic matrices, which provides critical input for network capacity planning and traffic engineering, will be contaminated by the dirty data created by troubles such us hardware/software/transmission. To solve this problem, a SNMP-data-self based sparse dirty data model and a sparse dirty data Processing Algorithm based on alternating projection are proposed here. The former occupies merit of no checkout between measurement of OD flows and SNMP data, the latter obtains norm minimized solution through alternating projection method. Comparing to other main algorithms, this method has lower measuring overhead and computing complexity. Moreover, the simulation experiment shows high accuracy.