Detail publikace

Evolution of Cache Replacement Policies to Track Heavy-hitter Flows

Originální název

Evolution of Cache Replacement Policies to Track Heavy-hitter Flows

Anglický název

Evolution of Cache Replacement Policies to Track Heavy-hitter Flows

Jazyk

en

Originální abstrakt

This paper presents a scheme to evolve fine-tuned/specialized replacement policy to keep track of heavy flows in network traffic. The evolved replacement policy provides a flow cache management mechanism to decide which flow states to preserve and which to expire. The observation shows that the well-known LRU and its modifications are not suitable replacement policies for network traffic stateful processing which focuses on heavy flows.  Therefore we introduce a general description of any replacement policy and let Genetic Algorithm to evolve novel replacement policy using this description.  The results shows that the evolved policy is more suitable for paradigm of heavy flow processing and monitoring. Moreover, our approach keeps state of heavy flows since the start-of-day. This is a significant difference to filtering approaches proposed in previous work which might many applications benefit from.

Anglický abstrakt

This paper presents a scheme to evolve fine-tuned/specialized replacement policy to keep track of heavy flows in network traffic. The evolved replacement policy provides a flow cache management mechanism to decide which flow states to preserve and which to expire. The observation shows that the well-known LRU and its modifications are not suitable replacement policies for network traffic stateful processing which focuses on heavy flows.  Therefore we introduce a general description of any replacement policy and let Genetic Algorithm to evolve novel replacement policy using this description.  The results shows that the evolved policy is more suitable for paradigm of heavy flow processing and monitoring. Moreover, our approach keeps state of heavy flows since the start-of-day. This is a significant difference to filtering approaches proposed in previous work which might many applications benefit from.

BibTex


@inproceedings{BUT34418,
  author="Martin {Žádník} and Marco {Canini}",
  title="Evolution of Cache Replacement Policies to Track Heavy-hitter Flows",
  annote="This paper presents a scheme to evolve fine-tuned/specialized replacement policy
to keep track of heavy flows in network traffic. The evolved replacement policy
provides a flow cache management mechanism to decide which flow states to
preserve and which to expire. The observation shows that the well-known LRU and
its modifications are not suitable replacement policies for network traffic
stateful processing which focuses on heavy flows.  Therefore we introduce
a general description of any replacement policy and let Genetic Algorithm to
evolve novel replacement policy using this description.  The results shows that
the evolved policy is more suitable for paradigm of heavy flow processing and
monitoring. Moreover, our approach keeps state of heavy flows since the
start-of-day. This is a significant difference to filtering approaches proposed
in previous work which might many applications benefit from.",
  address="Association for Computing Machinery",
  booktitle="Proceedings of the 6th ACM/IEEE Symposium on Architectures for Networking and Communications Systems",
  chapter="34418",
  edition="NEUVEDEN",
  howpublished="print",
  institution="Association for Computing Machinery",
  year="2010",
  month="october",
  pages="1--2",
  publisher="Association for Computing Machinery",
  type="conference paper"
}