Download Multi-Protocol Correlation: Data Record Analyses and Correlator
Transcript
CHAPTER 4. VALIDATION AND TESTING The reason for this correlation speed is largely because the correlator algorithm performs correlation by working with just the data in the memory. Type of CDR ISUP CDRs % of set to compare Processing Time ISUP/ISUP 82914 100% 1345 ms ISUP/ISUP 82914 100% 1541 ms ISUP/ISUP 82914 100% 1203 ms ISUP/ISUP 82914 100% 1225 ms ISUP/ISUP 82914 100% 1210 ms ISUP/ISUP 82914 100% 1214 ms ISUP/ISUP 82914 100% 1217 ms ISUP/ISUP 82914 100% 1183 ms ISUP/ISUP 82914 100% 1254 ms ISUP/ISUP 82914 100% 1251 ms Table 4.2: A dataset of ISUP Correlation Time performance on an Intel Core i5 2,26 Ghz processor running on a single thread using the Windows 7 x64 operating system. We also tried a different correlation process on a much larger sample with a more complex correlation process. In the case illustrated in Table 4.2 the correlation time is drastically increased to 1264 milliseconds in average, but that is still within acceptable limits considering correlation time. We believe that there are two main reasons for the increased time. The first reason being that the correlation pool is much larger, 82914 CDRs to correlate compared to the 10197 and 9760 from the previous correlation pools in table 4.1. The second reason is the actual correlation process itself; when correlating ISUP with ISUP we use a parameter called hold time. Our correlator compares them by seeing if they are within the user specified percentage. This process is much more CPU intensive because it requires division of numbers. The type of correlation performed in Table 4.2 is based on the basic ISUP - ISUP correlation case mentioned in subsection 3.3.1. The design of the algorithm is optimized for the correlation operations and the two dimensional linked list architecture used as the main data structure gives good 74