On unreliable computing systems when heavy-tails appear as a result of the recovery procedure
ACM SIGMETRICS Performance Evaluation Review - Special issue on the workshop on MAthematical performance Modeling And Analysis (MAMA 2005)
ACM SIGMETRICS Performance Evaluation Review
Dynamic packet fragmentation for wireless channels with failures
Proceedings of the 9th ACM international symposium on Mobile ad hoc networking and computing
Is ALOHA causing power law delays?
ITC20'07 Proceedings of the 20th international teletraffic conference on Managing traffic performance in converged networks
File fragmentation over an unreliable channel
INFOCOM'10 Proceedings of the 29th conference on Information communications
Transition from heavy to light tails in retransmission durations
INFOCOM'10 Proceedings of the 29th conference on Information communications
Retransmissions over correlated channels
ACM SIGMETRICS Performance Evaluation Review - Special issue on the 31st international symposium on computer performance, modeling, measurements and evaluation (IFIPWG 7.3 Performance 2013)
| Hi-index | 0.00 |
Retransmission-based failure recovery represents a primary approach in existing communication networks, on all protocol layers, that guarantees data delivery in the presence of channel failures. Contrary to the traditional belief that the number of retransmissions is geometrically distributed, a new phenomenon was discovered recently, which shows that retransmissions can cause long (-tailed) delays and instabilities even if all traffic and network characteristics are light-tailed, e.g., exponential or Gaussian. Since the preceding finding holds under the assumption that data sizes have infinite support, in this paper we investigate the practically important case of bounded data units 0≤ Lb≤ b. To this end, we provide an explicit and uniform characterization of the entire body of the retransmission distribution Pr[Nb n] in both n and b. This rigorous approximation clearly demonstrates the previously observed transition from power law distributions in the main body to exponential tails. The accuracy of our approximation is validated with a number of simulation experiments. Furthermore, the results highlight the importance of wisely determining the size of data units in order to accommodate the performance needs in retransmission-based systems. From a broader perspective, this study applies to any other system, e.g., computing, where restart mechanisms are employed after a job processing failure.