An investigation of packet reordering in TCP traces (extended abstract)
- Gabriel
- Anders A.
- Matthew S.
- Christopher L.
- Madhav V.
Recent research has highlighted the impact of packet reordering on network dynamics. Still, while much work has investigated the statistical properties of inter-packet arrival times of TCP traces, little effort has been devoted to obtaining a model of network traffic that incorporates sequence ID numbers as well. With the ultimate goal to develop such a joint model, they present results on the dynamics of packet reordering in a set of publicly available TCP traces recorded at the Network Research lab at UCLA. they investigate the scaling properties of the number of packet inversions. They propose a two-state model for the dynamics of sequence IDs based on pivots (defined as packets for which the received packet sequence has no gaps). This concept allows us to partition the trace into time epochs based on the presence or absence of reordering. Thys, they are able to identify and store patterns of reordering in the packet streams. Statistical tests provide a first-order validation of the model. Finally, they investigate the reordering patterns identified by their model from the standpoint of standard measures of presortedness of integer sequences. The methodology outlined in this paper enables regeneration of synthetic traces with inversion characteristics that are statistically similar to those of the original data. It is part of RESTORED, a network inference and analysis tool under development at Los Alamos National Laboratory.
- Research Organization:
- Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- USDOE
- OSTI ID:
- 977651
- Report Number(s):
- LA-UR-04-3552; TRN: US201012%%630
- Resource Relation:
- Conference: Submitted to: ACM Internet Measurement Conference, October 2004, Taormina
- Country of Publication:
- United States
- Language:
- English
Similar Records
Deep Packet/Flow Analysis using GPUs
Deep Packet/Flow Analysis using GPUs