S. Suri, G. Varghese, and G. Chandramenon. Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delay and Throughput Fairness. In Proceedings of INFOCOM '97, April 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....but it uses the expected completion times of previous computations, instead of computations to be scheduled, for scheduling. FRC s solution to the fairness problems is similar to several other approaches, such as virtual clock reset [20] time shift scheduling [2] and leap forward virtual clock [16]. Other rate based algorithms with suitable firewall protection can also be used in our framework. For system integration into a general purpose OS environment, we discussed issues such as priority inheritance and proxied scheduling. Diverse experimental results demonstrate the soundness and ....

S. Suri, G. Varghese, and G. Chandranmenon. Leap forward virtual clock: A new fair queueing scheme with guaranteed delays and throughput fairness. In Proc. IEEE Infocom 97, Kobe, Japan, April 1997.

....model. It can provide network delay bound for leaky bucket constrained traffic and has been intensively studied [3] 6] 9] However, because the fluid GPS is not practical, a class of PFQ algorithms has been proposed to emulate the fluid GPS to achieve the desired performance [7] 11] 19] [22]. All of them are based on maintaining a global function, referred to as either system virtual time or system potential, which tracks the progress of the GPS. This global function is used to compute a virtual Manuscript received July 14, 1998; revised February 15, 1999. H. J. Chao, X. Guo, and C. ....

....(FPGA) chips. Recently, the worst case fairness index (WFI) 17] has been introduced to measure how closely a packet by packet scheduler approximates the GPS system in a hierarchical scheduling environment. Another class of scheduling algorithms called shaper schedulers [13] 15] 17] 18] [22] has been proposed to achieve minimum WFI and have better worst case fairness properties. With these algorithms, when the server is picking the next packet to transmit, it chooses, among all the eligible packets, the one with the smallest time stamp. A packet is eligible if its virtual start time ....

S. Suri, G. Varghese, and G. Chandranmenon, "Leap forward virtual clock: A new fair queueing scheme with guaranteed delays and throughput fairness," in Proc. IEEE INFOCOM, Apr. 1997.

....and (iii) the complexity of sorting the timestamps of the eligible cells in order to select the one with minimum timestamp for the next service. The complexity of implementing the SEFF policy is a considerable burden when the scheduler s implementation is based on conventional priority queues [10, 11], since a worst case of O(V ) cells may become eligible at the same time. To solve this problem, Bennett et al. have introduced in [1] a simplified scheduling structure, referred to as the discrete rate approach, which can be used when the scheduler is only required to support a relatively small ....

....the discrete rate approach requires that the scheduler under consideration be worst case fair. However, the total implementation cost of the resulting discrete rate scheduler is not only dramatically lower than that of worst case fair schedulers implemented with conventional priority queues [10, 11], but is even competitive with the cost of non worst case fair schedulers implemented with other known techniques [12, 14] The competitive cost, together with the achievement of worst case fairness, explains the recent popularity of the discrete rate approach for the implementation of P RPS ....

[Article contains additional citation context not shown here]

S. Suri, G. Varghese and G. Chandranmenon, "Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness," Proc. INFOCOM '97, April 1997.

....the cell at the head of the corresponding queue, since this is the cell with the minimum starting potential among all cells in that queue. However, the complexity of implementing the SEFF policy is a considerable burden when the scheduler s implementation is based on conventional priority queues[11, 12], since a worst case of O(V ) cells may become eligible at the same time. Any P RPS scheduler using the SEFF selection policy achieves optimal delay bounds, is worstcase fair, and has a SFI very close to the theoretical lower bound in packet by packet servers [8, 11] The discrete rate approach, ....

....the discrete rate approach requires that the scheduler under consideration be worst case fair. However, the total implementation cost of the resulting discrete rate scheduler is not only dramatically lower than that of worst case fair schedulers implemented with conventional priority queues [11, 12], but is even competitive with the cost of non worst case fair schedulers implemented with other known techniques [13, 15] With the additional simplifications introduced in [2, 3] the total cost of the discrete rate scheduler can be further reduced, to the point of becoming comparable with the ....

S. Suri, G. Varghese and G. Chandranmenon, "Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness," Proc. INFOCOM '97, April 1997.

....as of O(log V ) see Section 2 below for a more extensive discussion of complexity issues) Several techniques have been proposed to reduce the cost of the sorting structure. For example, in case of software implementation, the approach in [6] explicitly applied to a GPS related scheduler in [7], reduces the complexity of the sorting task to O(log log V ) Other techniques have been specifically devised to simplify the hardware implementation of the sorting structure; they include calendar queues, such as the one used in [8] or the optimized one recently introduced in [9] and even ....

....P GPS. Starting Potential Fair Queueing (SPFQ) 4, 5] proposed as an improvement over FFQ, achieves the same delay bounds of P GPS and has fairness properties similar to those of P GPS, but its system potential function has complexity of O(log V ) The recent Leap Forward Virtual Clock (LFVC) [7] achieves very good delay bounds and has optimal fairness properties (more precisely, achieves worst case fairness [15] using a simple system potential function similar to the one of Virtual Clock, but requires an additional quarantine mechanism of worst case complexity of O(V log log V ) In ....

[Article contains additional citation context not shown here]

S. Suri, G. Varghese and G. Chandranmenon, "Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness," Proceedings of IEEE INFOCOM '97, April 1997.

....the transmission order in the idealized FFQ system, the priority value assigned to a packet is some function of its departure time in the FFQ system. Examples of sorted priority schemes are Virtual Clock (VC) 6] Self Clock Fair Queueing (SCFQ) 7] and Leap Forward Virtual Clock (LFVC) [8]. In the frame based approach, time is divided into frames and packets are entered into a frame without exceeding a maximum. An example of a frame based scheme is the Deficit Round Robin (DRR) method [9] Frame based methods are very scalable as the packet processing operations have constant time ....

....to flows, it does not provide any fairness guarantee. Packets from a flow that has been idle for a prolonged period of time will be assigned smaller virtual time stamp values and can receive for some period of time a larger than reserved share of service. The Leap Forward Virtual Clock [8] method solves the fairness problem in VC by temporarily moving oversubscribed flows into a low priority holding area. Only flows in the high priority area will receive service. A flow is oversubscribed if the difference between the virtual time stamps of the current packet of # and the system ....

S. Suri, G. Varghese and G. Chandranmenon, Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness, IEEE Infocom 1997.

....if at the cost of some decrease in performance guarantees or increase in the cost of other system components that are less critical for scalability purposes. One possible direction is to lower the cost of sorting by allowing some coarsening of the information to be sorted. This is the approach of [8], which achieves a reduction from log N to log log N in complexity, where N is the number of flows to be sorted. Another direction is to avoid reliance on sorting altogether as in the Rotating Priority Queue (RPQ) proposal of [10] This is the direction we pursue in this paper, where we take it to ....

....is required independently of any scheduling, if rate guarantees are to be provided. As a result, an approach that can provide rate guarantees by relying primarily on simple buffer management is attractive as it removes much of the complexity associated with scheduling. As with the schemes of [8] and [10] there is obviously some penalty associated with the simplification of providing rate guarantees only through buffer management. As we shall see in Section 2, this penalty is primarily in terms of looser delay guarantees to individual flows and an increase in the amount of buffer space ....

S. Suri, G. Varghese, and G. Chandranmenon. Leap forward virtual clock: A new fair queueing scheme with guaranteed dealy and throughput fairness. In Proceedings of INFOCOM, pages 558--566, Kobe, Japan, April 1997.

....provide per flow rate or delay guarantees. In order to preserve the flexibility offered by per flow management, recent research effort has focussed on reducing the complexity and increasing the efficiency of sorting and perflow classification. Sorting complexity of O(loglog n) has been reported in [25]. On the other hand, sorting is completely avoided in [8, 26] In [8] simple buffer management schemes are used to provide rate guarantees however, the buffer space is mostly statically partitioned, which results in poor resource utilization. It is widely believed that per flow classification ....

S. Suri, G. Varghese, and G. Chandramenon. Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delay and Throughput Fairness. In Proceedings of INFOCOM '97, April 1997.

....on the scalability of the QoS provision mechanism. Packet admission decisions made by buffer management methods typically require only a constant amount of processing and state information [5] Packet scheduling decisions can be computationally intense. For example, WFQlike schedulers [2] 3] [4] maintain an ordered queue and have non constant time decision operations. Packet scheduling affects the queueing delay. Queueing delay can be significant in lower speed networks, however, it is relatively small in very high speed links. For example, an OC48 link at 2.4 Gbits sec can empty a 1 ....

....in [6] and [2] respectively. The WFQ scheduler requires O( time to schedule a packet, where is the number of active flows. More efficient scheduling algorithms that approximate the WFQ scheduler have been developed, for example, Self clock Fair Queueing [3] and Leap Forward Virtual Clock [4]. In the remainder of this paper, we will present the PS buffer management method and show that it can provide rate guarantees to leaky bucket constrained flows. We will also compare the buffer reservation, throughput, delay and jitter performance of PS with the method in [5] which can provide ....

S. Suri, G. Varghese and G. Chandranmenon, Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness, IEEE Infocom 1997, pp. 557--565, 1997.

....but it uses the expected completion times of previous computations, instead of computations to be scheduled, for scheduling. FRC s solution to the fairness problems is similar to several other approaches, such as virtual clock reset [24] time shift scheduling [2] and leap forward virtual clock [20]. Other rate based algorithms with suitable firewall protection can also be used in our framework. To the best of our knowledge, ours is the first implementation in a general purpose OS environment of the proposed heterogeneous services architecture. Issues of system integration in such an ....

S. Suri, G. Varghese, and G. Chandranmenon. Leap forward virtual clock: A new fair queueing scheme with guaranteed delays and throughput fairness. In Proc. IEEE Infocom 97, Kobe, Japan, April 1997.

No context found.

S. Suri, G. Varghese, and G. Chandramenon. Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delay and Throughput Fairness. In Proceedings of INFOCOM '97, April 1997.

No context found.

S. Suri, G. Varghese and G. Chandranmenon, Leap Forward Virtual Clock: A New Fair Queueing Scheme with Guaranteed Delays and Throughput Fairness, IEEE Infocom 1997.

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