Journal of the ACM (JACM)
Competitve buffer management for shared-memory switches
Proceedings of the thirteenth annual ACM symposium on Parallel algorithms and architectures
Buffer overflow management in QoS switches
STOC '01 Proceedings of the thirty-third annual ACM symposium on Theory of computing
Nearly optimal FIFO buffer management for DiffServ
Proceedings of the twenty-first annual symposium on Principles of distributed computing
Dynamic routing on networks with fixed-size buffers
SODA '03 Proceedings of the fourteenth annual ACM-SIAM symposium on Discrete algorithms
Management of multi-queue switches in QoS networks
Proceedings of the thirty-fifth annual ACM symposium on Theory of computing
Packet dropping policies for ATM and IP networks
IEEE Communications Surveys & Tutorials
The network as a storage device: dynamic routing with bounded buffers
APPROX'05/RANDOM'05 Proceedings of the 8th international workshop on Approximation, Randomization and Combinatorial Optimization Problems, and Proceedings of the 9th international conference on Randamization and Computation: algorithms and techniques
| Hi-index | 0.00 |
Consider an Internet service provider (ISP), or a corporate intranet, that connects a large number of users with the Internet backbone using an "uplink." Within such a system, consider the traffic oriented towards the uplink, namely the streams whose start points are the local users and whose destination is outside the local domain. These streams are merged by a network that consists of merge nodes, typically arranged in a tree topology whose root is directly connected to the uplink. Without loss of generality, we may assume that the bandwidth of the link emanating from a merge node is less than the sum of bandwidths of incoming links (otherwise, we can assume that the incoming links are connected directly to the next node up). Hence, when all users inject data at maximum local speed, packets will eventually be discarded. A very effective way to mitigate some of the losses due to temporary overloads is to equip the merge nodes with buffers, that can absorb transient bursts by storing incoming packets while the outgoing link is busy. The merge nodes are controlled by local on-line buffer management algorithms whose job is to decide which packets to forward and which to drop so as to minimize the damage in case of an overflow.