V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... of MWM type schedulers can be found in, for example, 7, 8, 9, 10, 11] Framebased schedulers were presented in [19, 20] If the switch fabric has an internal speedup of 2 then it is known that it can emulate output queued switches (in which there is no contention at the inputs) 21, 22, 23] In [24], an algorithm is presented whose aim is to track an idealized fluid policy. If the switch is sufficiently underloaded then tight delay bounds can be achieved. In [25] it is shown that if the total load on any input or output is at most one quarter of the link rate, then it is possible to serve ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

.... of MWM type schedulers can be found in, for example, 7, 8, 9, 10, 11] Framebased schedulers were presented in [19, 20] If the switch fabric has an internal speedup of then it is known that it can emulate output queued switches (in which there is no contention at the inputs) 21, 22, 23] In [24], an algorithm is presented whose aim is to track an idealized fluid policy. If the switch is sufficiently underloaded then tight delay bounds can be achieved. In [25] it is shown that if the total load on any input or output is at most one quarter of the link rate, then it is possible to serve ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

.... can be found in, for example, 7] 8] 9] 10] 11] Some frame based schedulers were presented in [19] 20] If the switch fabric has an internal speedup of then it is known that it can emulate output queued switches (in which there is no contention at the inputs) 21] 22] 23] In [24], an algorithm is presented whose aim is to track an idealized fluid policy. If the switch is sufficiently underloaded then tight delay bounds can be achieved. In [25] it is shown that if the total load on any input or output is at most one quarter of the link rate, then it is possible to serve ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

.... can be found in, for example, 7] 8] 9] 10] 11] Some frame based schedulers were presented in [19] 20] If the switch fabric has an internal speedup of then it is known that it can emulate output queued switches (in which there is no contention at the inputs) 21] 22] 23] In [24], an algorithm is presented whose aim is to track an idealized fluid policy. If the switch is sufficiently underloaded then tight delay bounds can be achieved. In [25] it is shown that if the total load on any input or output is at most one quarter of the link rate, then it is possible to serve ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

.... can be found in, for example, 7] 8] 9] 10] 11] Some frame based schedulers were presented in [19] 20] If the switch fabric has an internal speedup of 2 then it is known that it can emulate output queued switches (in which there is no contention at the inputs) 21] 22] 23] In [24], an algorithm is presented whose aim is to track an idealized fluid policy. If the switch is sufficiently underloaded then tight delay bounds can be achieved. In [25] it is shown that if the total load on any input or output is at most one quarter of the link rate, then it is possible to serve ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

.... ports (LPF policy) 2] To the best of our knowledge, instead, no general result exists on the performance of pure IQ switches dealing with multiple traffic classes; only heuristic scheduling algorithms supporting multiple traffic classes were proposed in the recent literature [12] 13] 14] [15], and their performance was assessed by simulation for a limited number of traffic patterns. A wider set of results are known for CIOQ switch architectures. CIOQ switch with speed up equal to 2 have been proved to be able to exactly emulate OQ switches implementing any monotonic work conserving ....

V.Tabatabaee, L.Georgiadis, L.Tassiulas, "QoS provisioning and tracking fluid policies in input queueing switches", INFOCOM

....the recent advances in gigabit optical networking, the switching fabric may not be fast enough, and input buffering may be employed. Unfortunately, in such a case, tracking a fluid scheduling policy (such as GPS) with a non anticipative packet scheduling policy is still an open problem (see, e.g. [33] and references therein for a link to the multi periodic TDMA satellite scheduling problem) The importance of input output buffering notwithstanding, the implementation cost of PFQ Deterministic Time Varying Packet Fair Queueing 83 Figure 7. Bandwidth allocation under SCED. Figure 8. Bandwidth ....

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS Provisioning and Tracking Fluid Policies in Input Queueing Switches," in Proc. of INFOCOM'

.... ports (LPF policy) 2] To the best of our knowledge, instead, no general result exists on the performance of pure IQ switches dealing with multiple traffic classes; only heuristic scheduling algorithms supporting multiple traffic classes were proposed in the recent literature [12] 13] 14] [15], and their performance was assessed by simulation for a limited number of traffic patterns. A wider set of results are known for CIOQ switch architectures. CIOQ switch with speed up equal to 2 have been proved to be able to exactly emulate OQ switches implementing any monotonic work conserving ....

V.Tabatabaee, L.Georgiadis, L.Tassiulas, "QoS provisioning and tracking fluid policies in input queueing switches", INFOCOM

....to the case when P=m(i; j) is an integer for each (i; j) in the present formulation. Giles provided a constructive proof of this conjecture in the case when each period P=m(i; j) evenly divides all longer periods[8] A slightly more general model with time dependent service rates is considered in [9], and a scheduling algorithm is given for the case N = 2. The rest of the paper is organized as follows. Section II involves the formulation of the scheduling problem, and gives a characterization of its solutions. This characterization is used in Section III to obtain a number of heuristic ....

....1.0. For more general period, the algorithm provides a feasible schedule if link utility is less than 1 4. Another greedy algorithm which provides a schedule for arbitrary periods for a switch with link utility less than 1 14 is also provided in [8] Finally, a heuristic algorithm proposed in [9] provides similar performance with the proposed algorithms in this paper in terms of the percentage of missed deadlines however its performance was obtained for average link utility of 0.92. Additionally, the algorithm proposed in [9] is far more complex due to its critical nodes and links ....

[Article contains additional citation context not shown here]

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, QoS provisioning and tracking fluid policies in input queuing switches, Proc. IEEE INFOCOM'00, Mar. 2000.

....each (i; j) in the present formulation. The sequence of deadlines has period P=m(i; j) in this case, and [2] provided a constructive proof of the conjecture when each period evenly divides all longer periods. For general periods, the provided schedules apply if line utilizations do not exceed 1 4. [3] identified a scheduling algorithm to meet deadlines in 2 2 Theta 2 switches with time varying connection rates, and numerically studied some algorithms for general switch sizes under the traffic model of [1] A service schedule for one time slot of the switch fabric is represented by a ....

....time. 5 3 Conclusions The scheduling algorithms presented here have favorable features compared to previous related work. Their performances improve with increasing switch size and schedule length in contrast to those proposed in [1] and they work well under full line utilization whereas [2, 3] report similar performance for smaller utilization. It is reasonable to expect that in typical applications connections are set up and torn down on a much slower time scale compared to the unit time slot, so the algorithms can be implemented online to support such dynamic scenarios. It appears ....

[Article contains additional citation context not shown here]

V. Tabatabaee, L. Georgiadis and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," Proc. IEEE INFOCOM'

....matching algorithms that function based on the port weight rather than link weights. Weight of a port is total weight of all links connected to it. LPF [9] is one of these algorithms, where we consider the backlogged traffic in every port as the weight of that node. Maximum node matching (MNM) [14] is another example that weights are amount of service that scheduler owes every link, according to the reserved rate of that port. The only difference between the MNM and LPF is in the weight function that is used for links, but they have the same complexity. It is proved in [14] that the ....

....matching (MNM) 14] is another example that weights are amount of service that scheduler owes every link, according to the reserved rate of that port. The only difference between the MNM and LPF is in the weight function that is used for links, but they have the same complexity. It is proved in [14] that the obtained matching under MNM is a min max matching, and it has the maximum lexicographic ordering. In other words, it is impossible to add a new port to the matching, without removing a higher weight node from the matching. This is also true for LPF for the corresponding weight vectors. ....

[Article contains additional citation context not shown here]

V. Tabatabaee, L. Georgiadis, L. Tassiulas, " QoS Provisioning and Tracking Fluid Policies in Input Queueing Switches", IEEE Transaction on Networking, v.9, no.5, pp. 605-617, Oct. 2001.

No context found.

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queuing switches," IEEE/ACM Trans. on Networking, vol. 9, no. 5, October 2001.

No context found.

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queueing switches," in INFOCOM, 2000, pp. 1624--1633.

No context found.

V. Tabatabaee, L. Georgiadis, and L. Tassiulas, "QoS provisioning and tracking fluid policies in input queueing switches," in INFOCOM, 2000, pp. 1624--1633.

No context found.

V.Tabatabaee, L.Georgiadis, L.Tassiulas, "QoS provisioning and tracking fluid policies in input queueing switches", INFOCOM

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