| Wu chang Feng, Dilip Kandlur, Debanjan Saha, and Kang Shin. Blue: A New Class of Active Queue Management Algorithms. Technical Report CSE-TR-387-99, University of Michigan, April 1999. |
....for this paper is comparing RED byte packet RED can mark packets by either dropping the packet or using an explicit signal (e.g. RFB01] This assumes the gentle variant of RED [Flo00] which we used in our simulations. variants with alternate queueing strategies (e.g. droptail, BLUE [cFKSS99], REM [ALLY01] FRED [LM97] etc. We do, however, believe that investigating biases in other queueing disciplines would be useful future work. The remainder of this paper is organized as follows. x 2 outlines the simulation environment we use to test the performance of the RED variants. x 3 ....
Wu chang Feng, Dilip Kandlur, Debanjan Saha, and Kang Shin. Blue: A New Class of Active Queue Management Algorithms. Technical Report CSE-TR-387-99, University of Michigan, April 1999.
....this sample. The dropping probability of the incoming packet is determined by the number of packets in the sample from the same flow. This is a simultaneous but independent work that is still unpublished. Our comments are based on discussions with one of the authors. Stochastic Fair Blue (SFB) [FKSS99] does not use per flow state, but relies on multiple levels of hashing to identify highbandwidth flows. As the authors state in their paper, the scheme works well when there are only a few high bandwidth flows. In the presence of multiple high bandwidth flows SFB could end up punishing even the ....
Wu-Chang Feng, Dilip D. Kandlur, Debanjan Saha, and Kang Shin. Blue: A New Class of Active Queue Management Algorithms. Technical Report CSE-TR-387-99, University of Michigan, April 1999.
....at the router. FRED s fair allocation of buffers can yield very different fairness properties from a fair allocation of bandwidth [SSZ98] In addition, the results obtained by FRED are not predictable, as they depend on the packet arrival times of the individual flows. Stochastic Fair Blue (SFB) FKSS99] does not use per flow state to achieve fairer allocations but relies on multiple levels of hashing to identify high bandwidth flows. As the authors state in their paper, the scheme works well when there are only a few high bandwidth flows. In the presence of multiple high bandwidth flows it ends ....
....about a flow in the history means that a flow is high bandwidth. ffl Bloom Filters or Hashing: Incoming flows could be hashed into bins at one or more levels, and the bins with more hits could be used to identify high bandwidth flows. The identification mechanism in Stochastic Fair 5 Blue [FKSS99] is an example of this approach. A hashing based approach avoids the memory and processing required for full per flow state, with the risk of incorrectly identifying low bandwidth flows mapped to the same bins as high bandwidth flows. RED PD uses the RED drop history to identify highbandwidth ....
Wu-Chang Feng, Dilip D. Kandlur, Debanjan Saha, and Kang Shin. Blue: A New Class of Active Queue Management Algorithms. Technical Report CSE-TR-387-99, University of Michigan, April 1999.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC