| N. McKeown, \Scheduling Algorithms for Input-Queued Cell Switches," Ph.D. Thesis, University of California at Berkeley, 1995. |
....this paper we discuss and compare a number of such proposals in the context of packet switches operating on fixed size data units. The term cell switches will be used, although the discussed switch architectures need not refer to ATM. The considered IQ cell switch architecture proposals are: iSLIP [5], 6] iLQF [5] 7] iOCF [5] 7] iLPF [3] 2DRR [8] RC [9] MUCS [10] RPA [11] 12] and iZIP [13] a novel proposal. II. A CLOSER LOOK AT IQ CELL SWITCH ARCHITECTURES We consider IQ switch architectures with synchronous operations, where time is slotted, and where fixed size data units are ....
....and compare a number of such proposals in the context of packet switches operating on fixed size data units. The term cell switches will be used, although the discussed switch architectures need not refer to ATM. The considered IQ cell switch architecture proposals are: iSLIP [5] 6] iLQF [5], 7] iOCF [5] 7] iLPF [3] 2DRR [8] RC [9] MUCS [10] RPA [11] 12] and iZIP [13] a novel proposal. II. A CLOSER LOOK AT IQ CELL SWITCH ARCHITECTURES We consider IQ switch architectures with synchronous operations, where time is slotted, and where fixed size data units are transferred in ....
[Article contains additional citation context not shown here]
McKeown N., "Scheduling Algorithms for Input-Queued Cell Switches", Ph.D.Thesis, University of California at Berkeley, 1995
.... parallel buffered switches, capable of performing parallel read write, have received a lot of attention recently (see e.g. 15] 16] and references therein) Traditionally, the study of parallel buffered switches is limited to the (singlestage) input buffered crossbar switch (see e.g. 17] [22], 27] 23] 19] 9] 24] 28] 14] 20] 1] 21] where each input has a segregated buffer. In such a switch, time is slotted and synchronized so that packets in different input buffers can be read out simultaneously within a time slot. There This research is supported in part by the ....
N. McKeown, Scheduling algorithms for input-queued cell switches, PhD Thesis. University of California at Berkeley, 1995.
....to avoid contention at inputs and thus guarantee high performance. Considerable work has been done on improving performance using the matching algorithms. It has been proved that by using some maximum weight matching algorithm 100 throughput can be reached for independent arrivals [11] 12] [13]. But maximum match is not practical to implement in hardware due to its complexity, and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [4] 9] 14] but maximal matching algorithm cannot achieve as high a throughput as ....
....complexity, and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [4] 9] 14] but maximal matching algorithm cannot achieve as high a throughput as maximum matching algorithms. Iterative algorithms as PIM [1] and iSLIP [10] [13], use multiple iterations to converge on a maximal matching. One way to improve the performance of a VOQ switch is to increase the speedup of the switch fabric. In [5] 9] 17] a Combined Input and Output Queuing (CIOQ) switch with some matching algorithms have been proved to reach up to 100 ....
[Article contains additional citation context not shown here]
N. McKeown, "Scheduling algorithms for input-queued cell switches", Ph.D. Thesis, UC Berkeley, May 1995.
....Six MetroTech Center, Brooklyn, NY, 11201 3840, USA. Fax: 7182603906 Considerable work has been done on scheduling algorithms for VOQ switches. It has been proved that by using a maximum weight matching algorithm #### throughput can be reached for i.i.d. arrivals (uniform or nonuniform) 3] 4] [5]. But maximum match is not practical to implement in hardware due to its complexity, and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [6] 7] 8] but maximal matching algorithm cannot achieve as high a throughput as ....
....and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [6] 7] 8] but maximal matching algorithm cannot achieve as high a throughput as maximum matching algorithms. Iterative algorithms such as PIM [9] and iSLIP [10] [5], use multiple iterations to converge on a maximal matching. The Dual Round Robin Matching (DRRM) switch [11] 12] builds and improves on the ideas incorporated in iSLIP. It has been proven that DRRM can achieve #### throughput under i.i.d. and uniform traffic [12] Furthermore, the DRRM scheme ....
N. McKeown, "Scheduling algorithms for input-queued cell switches", Ph.D. Thesis, UC Berkeley, May 1995.
....Science Foundation under grants ANI0081527 and ANI0081357 . Considerable work has been done on scheduling algorithms for VOQ switches. It has been proved that by using a maximum weight matching algorithm throughput can be reached for i.i.d. arrivals (uniform or nonuniform) 3] 4] 5] [6]. But maximum weight matching is not practical to implement in hardware due to its complexity, and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [7] 8] 9] but maximal matching algorithm cannot achieve as high a ....
....and may not guarantee fairness and quality of service. A number of practical maximal matching algorithms have been proposed [7] 8] 9] but maximal matching algorithm cannot achieve as high a throughput as maximum matching algorithms. Iterative algorithms such as PIM [10] and iSLIP [11] [6], use multiple iterations to converge on a maximal matching. The Dual Round Robin Matching (DRRM) switch [12] 13] builds and improves on the ideas incorporated in iSLIP. It has been proven that DRRM can achieve throughput under i.i.d. and uniform traffic [13] Furthermore, the DRRM ....
N. McKeown, "Scheduling algorithms for input-queued cell switches", Ph.D. Thesis, UC Berkeley, May 1995.
.... that there exist scheduling algorithms able to exploit a throughput of the 100 and also to avoid starvation of cells ( 14] However, such algorithms have several drawbacks, that can be so classified: Complexity: an optimal scheduling can be found solving a matching problem on bipartite graphs ([15]) or finding a decomposition of stochastic matrices (see [16] for the switching case) The weak point of these approaches is their complexity; the best known matching algorithm runs in O(N 2 log2 (N) time in the worst case ( 17] while the second method has been proved useful to implement, at ....
.... is their complexity; the best known matching algorithm runs in O(N 2 log2 (N) time in the worst case ( 17] while the second method has been proved useful to implement, at most, 4 x 4 switches ( 18] Throughput: with approximate algorithms it is possible to overcome the com plexity problem ([2, 15, 19, 13]) Behind this approach there are good simulation results ( 15, 20] and a proof that, if traffic reaches a steady state (i.e. there is always a cell that must be sent to every output) the behavior of these algorithms is optimal ( 13] But with bursty franc ( 21] such a stability is never ....
[Article contains additional citation context not shown here]
N. McKeown. Scheduling algorithms for input queued cell switches. PhD thesis, University of California at Berkeley, 1995.
....Another version of input queuing is virtual output queuing (VOQ) 1] VOQ combines the advantages of input queuing and output queuing. It has a switch like in input queuing and has the link utilization close to that of output queuing; 100 link utilization can still be achieved, when N is large [11]. As for output queuing, there are N queues. For every input i there are N queues Q(i, o) one for each output o, see Figure 2(b) Typically the set of N queues at each input port of a VOQ router are mapped onto a single RAM. However, for NoCs we strive at a small router and therefore we ....
....3(c) is a match of Figure 3(a) u 1 u 2 v 1 v 2 (c) accept u 1 u 2 v 1 v 2 u 1 u 2 v 1 v 2 (a) request (b) grant Figure 3. The three stages of a schedule iteration. Matching can be done optimally, but because of time complexity and fairness, a non optimal algorithm is preferred [11]. Our matching algorithm is iterative and one iteration has three stages, illustrated by an example in Figure 3 for N = 4. In the first stage, see Figure 3(a) every non empty queue Q(i, o) requests access to output port o from input port i. In the second stage, see Figure 3(b) every output port ....
[Article contains additional citation context not shown here]
N. McKeown. Scheduling Algorithms for Input-Queued Cell Switches. PhD thesis, Univ. of California, Berkeley, 1995.
....and the computation of weights used in the algorithms of [4] 11] 21] requires too much information to be communicated between inputs and outputs. Implementation considerations have therefore seen the proposal of a number of practicable scheduling algorithms; notably, iSLIP [15] iLQF [14], RPA [1] MUCS [6] and WFA [23] However, these algorithms perform poorly compared to MWM when the input traffic is non uniform: they induce very large delays and their throughput can be less than 100 . More recently, some particularly simple to implement scheduling algorithms have been proposed ....
McKeown N., "Scheduling algorithms for input-queued cell switches", Ph.D. Thesis, University of California at Berkeley, 1995
.... parallel buffered switches, capable of performing parallel read write, have received a lot of attention recently (see e.g. 15] 16] and references therein) Traditionally, the study of parallel buffered switches is limited to the (singlestage) input buffered crossbar switch (see e.g. 17] [22], 27] 23] 19] 9] 24] 28] 14] 20] 1] 21] where each input has a segregated buffer. In such a switch, time is slotted and synchronized so that packets in different input buffers can be read out simultaneously within a time slot. There This research is supported in part by the ....
N. McKeown, "Scheduling algorithms for input-queued cell switches," PhD Thesis. University of California at Berkeley, 1995.
....law of large numbers and that it does not oversubscribe any link in the network. I. INTRODUCTION AND PREVIOUS WORK Input Queueing (IQ) switch architectures have become an attractive architectural solution for the design of large size and high capacity packet switches. Anderson [1] and McKeown [2] showed in their seminal works that the negative effects of Head of the Line (HoL) blocking on performance can be reduced or completely eliminated by adopting per destination queueing (also called Virtual Output Queueing VOQ) at input cards, and by controlling the switch operations with a ....
N.McKeown, Scheduling Algorithms for Input-Queued Cell Switches, Ph.D. Thesis, Un. of California at Berkeley, 1995.
.... each slot, a scheduling algorithm decides the configuration of the switch by finding a matching on a bipartite graph (described below) A number of different techniques have been used for finding such a matching, for example, using neural networks [2] 4] 22] or iterative algorithms [1] 14] [15]. These algo 0090 6778 99 10.00 1999 IEEE rithms were designed to give high throughput while remaining simple to implement in hardware. When traffic is uniform, these algorithms perform well ( 90 throughput) The iSLIP algorithm [14] 15] for example, has been demonstrated using simulation ....
....[2] 4] 22] or iterative algorithms [1] 14] 15] These algo 0090 6778 99 10.00 1999 IEEE rithms were designed to give high throughput while remaining simple to implement in hardware. When traffic is uniform, these algorithms perform well ( 90 throughput) The iSLIP algorithm [14] [15], for example, has been demonstrated using simulation to achieve 100 throughput when the traffic is independent and uniform. However, all of these algorithms perform less well and are unable to sustain a throughput of 100 when traffic is nonuniform. It is worth asking the question: What is the ....
[Article contains additional citation context not shown here]
N. McKeown, "Scheduling algorithms for input-queued cell switches," Ph.D. dissertation, Univ. California at Berkeley, 1995.
No context found.
N. McKeown, \Scheduling Algorithms for Input-Queued Cell Switches," Ph.D. Thesis, University of California at Berkeley, 1995.
No context found.
+ McKeown, N+ ~1995!+ Scheduling algorithms for input-queued cell switches+ Ph+D+ thesis, University of California, Berkeley+
No context found.
N. McKeown, "Scheduling Algorithms for Input-Queued Cell Switches," Ph. D. Thesis, University of California at Berkeley, 1995.
No context found.
N. McKeown, "Scheduling algorithms for input-queued cell switches," Ph.D. dissertation, Univ. of California, Berkeley, 1995.
No context found.
N. McKeown, "Scheduling algorithms for input-queued cell switches," PhD Thesis. University of California at Berkeley, 1995.
No context found.
N. McKeown, Scheduling Algorithms for Input-Queued Cell Switches, (Ph. D. Thesis, University of California, Berkeley, 1995).
No context found.
N. Mckeown, "Scheduling Algorithms for Input-queued Cell Switches," Ph. D. dissertation, Univ. California at Berkeley, 1995.
No context found.
Mckeown, W.N.: `Scheduling algorithms for input-queued cell switches'. PhD thesis, University of California, Berkeley, CA, USA,
No context found.
N.McKeown, Scheduling algorithms for input-queued cell switches, Ph.D. Thesis, Un. of California at Berkeley, 1995.
No context found.
N. McKeown. Scheduling algorithms for input-queued cell switches, 1995.
No context found.
N. McKeown, \Scheduling algorithms for input-queued cell switches," PhD Thesis. University of California at Berkeley, 1995.
No context found.
N. McKeown, " Scheduling Algorithms for Inputqueued Cell Switches," Ph.D. thesis, UC Berkeley, May 1995.
No context found.
N. McKeown, "Scheduling algorithms for input-queued cell switches," PhD Thesis, University of California, Berkeley, May 1995.
No context found.
N. McKeown, "Scheduling algorithms for input-queued cell switches", PhD Thesis, University of California, Berkeley, May 1995.
First 50 documents Next 50
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