Scopul nostru este sprijinirea şi promovarea cercetării ştiinţifice şi facilitarea comunicării între cercetătorii români din întreaga lume.
Autori: Gabriel Istrate, Anders Hansson and Guanhua Yan
Editorial: IEEE Computer Society Press, ISBN 0-7695-2622-5, Proceedings of the Second International Conference on Networking and Services (ICNS'06), Workshop on Internet Packet Dynamics, San Jose CA, 2006.
Deciding what makes a packet reordering metric meaningful is a problem that has attracted significant interest in the computer networking community (See for instance the IETF Internet draft “Packet Reordering metric for IPPM”), but still lacks a universally accepted solution. We add to this discussion by the definition and investigation of some theoretical concepts illustrating the following simple points:
1. The notion of two traces being “the same modulo unimportant details” should be precisely specified using similarity relations.
2. A metric that is inconsistent, i.e. gives different values on two similar TCP traces, should not be regarded as meaningful. Surprisingly, such a natural restriction seems not to have been considered before.
3. We illustrate the concept of consistency in the context of RESTORED, an approach to semantic modeling and compression of TCP traces. In particular, we discuss the consistency of two metrics defined by Jayasumana et al. (LCN’02, IFIP NETWORKING’05, etc) with respect to behavioral equivalence and the other related similarity notions that were motivated by RESTORED (Istrate et al., IFIP NETWORKING’06, etc).
4. We also discuss the use of reorder invariants for network traffic as a way to make the above concepts more realistic. We present one metric from the sorting algorithms literature, called Sorted Up Sequences (SUS). We present experimental evidence (arising from the analysis of real network data, as well as ns-2 simulations) that SUS is indeed a reordering invariant.
Cuvinte cheie: TCP, packet reordering, reordering metrics, consistency, reorder invariants