Storage space saving and error avoidance of passive delay variation measurement employing a hash-based technique for IP networks

Delay variation of packets is a significant performance measure in IP networks. Both active and passive measurements can evaluate delay variation. Among these, passive measurement is more advantageous because it can monitor the performance experienced by users without disturbing network load. Meanwhile, a shortcoming of using passive measurement is an extremely large storage space required to record packets and timestamps. This problem is solved by feeding a portion of each packet to a hash function as its input and recording the output to identify the packet. Since the data length of the hash function output is very short, such an approach effectively decreases the storage space. However, this approach still needs a considerably large storage space to record time-stamps. Additionally, the collisions among hash output values may bring unacceptable measurement errors. This paper proposes a method to decrease the storage space required to record timestamps for hash-based delay variation measurement. This decrease is achieved by a variable length coding of timestamps. The paper also presents methods to avoid measurement errors brought by hash value collisions. Through experiments, it is shown that the proposed techniques correctly measure delay variation as well as effectively reduce the storage space. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 90(3): 21–32, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.20328