| S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2(3):157--173, September 1991. |
....3: State diagram of demotion and promotion within three colors. The leaky bucket is a deterministic ow control network element that can be used as a trac marker. Like the drop tail queue, a simple leaky bucket demotes all packets that arrive when there are no tokens available. As argued in [16], much of the Internet trac is highly periodic, either because of periodic sources (e.g. real time audio or video) or because window ow control protocols have a periodic cycle equal to the connection round trip time (e.g. a network bandwidth limited TCP bulk data transfer) This phase e ect ....
....the aggregate ow, i.e. every other packet must be demoted. Then all the packets from one ow will be demoted while all the packets from the other ow will remain una ected. Phase e ect could also bring about unfairness in promotion. Detail discussions on the phase e ect have been reported in [16]. Introducing randomness in the packet selection process of the ow control mechanism could solve the problem. An example is the Random Early Detection (RED) gateway [17] that reduces the unfairness of the drop tail queue. We apply a similar concept to the leaky bucket marker by introducing ....
[Article contains additional citation context not shown here]
S. Floyd and V. Jacobson, \On Trac Phase Eects in Packet-Switched Gateways," Internetworking: Research and Experience, V.3 N.3, Sep 1992, p.115-156.
....with different priorities. A queue with a higher priority is always served first. Priority queueing is the simplest form of preferential queueing. However, low priority traffic easily starves unless there is a mechanism to regulate high priority traffic. WFQ WFQ (Weighted Fair Queueing) [11, 4, 8] is a discipline that assigns an independent queue for each flow. WFQ can provide fair bandwidth allocation in times of congestion, and protects a flow from other flows. A weight can be assigned to each queue to give a different proportion of the network capacity. SFQ SFQ (Stochastic Fairness ....
Srinivasan Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2:157--173, September 1991.
....1, with drop tail queueing at node n 1 , leads to ow synchronization, i.e. the sending rates of all the ows exhibit in phase periodic variations. This produces undesirably large variations of the round trip time and poor utilization of the queue. This type of behavior has been observed before [5] and actually led to the development of Random Early Detection Drop active queuing [6] To the best of our knowledge, this is the rst time that the synchronization phenomena is theoretically explained. 2 Hybrid Model for Congestion Control In this paper, we consider Reno congestion control. We ....
S. Floyd and V. Jacobson, \On trac phase eects in packet-switched gateways," Internetworking: Research and Experience, vol. 3, pp. 115-116, Sept. 1992.
.... basic scheduler operational principle is the one formally expressed in formula (1) This principle determined the scheduler design which was, nevertheless, influenced by some lessons learned with the study and test of other disciplines [27] as, for instance, the weighted fair queueing discipline [15], 8] Early on, it was realised that it was not possible to use a pure work conserving packet scheduler discipline in the platforms available at LCT UC. In fact, the dynamics of the systems in use tends to serialise the appearance of packets in queues [27] 29] inhibiting any capacity of such ....
Srinivasan Keshav, On the efficient implementation of fair queuing, Internetworking: Research and Experience, September 1991.
....periods of congestion. For example, Random Early Detection (RED) gateways [12] can achieve this by marking packets with a certain probability when the queue length exceeds a threshold. Another method to ensure proper network behavior is isolation of ows using a mechanism like fair queueing [8] so that conforming ows are not penalized by an aggressive sender. Mechanisms such as integrated services [6] and di erentiated services [20] also attempt to provide guarantees per ow or trac class and achieve a similar e ect. However, all these schemes rely on modi cations to routers which is a ....
A. Demers, S. Keshav, and S. Shenker. Analysis and Simulations of a FairQueueing Algorithm. Internetworking: Research and Experience, V(17):3-26, 1990.
.... An abstraction called virtual time helps us here it is a monotonic function of time, with the useful property that it maps each packet s starting and finishing times into values that can be determined as soon as the packet has arrived [2] This makes it possible to implement FQ quite efficiently [10]. It is shown in [5] that starting times can be used instead of finishing times, and that in either case the error in approximating PS remains bounded. As defined so far, PS and FQ both assume an equal rights policy for all of their users, in that all users are to be serviced at equal rates. A ....
Keshav, S. On the Efficient Implementation of Fair Queueing. Internetworking: Research and Experience 2 (1991), 157-173.
....ts are shown in Table 2. A phenomenological explanation of the loss burst nonlinearity is not immediately obvious, but we propose the following theory. It has been noted that modest loss bursts for a stream of packets can be caused by the drop tail queueing mechanism that many routers employ [9]. In particular, when routers are overloaded, they will drop all packets that arrive at the tails of their busy queues until the congestion is resolved. A packet ow that passes through this router may lose a number of consecutive packets. This behavior does seem to explain the loss bursts of 1 to ....
S. Floyd and V. Jacobson, \On trac phase-eects in packet-switched gateways," Internetworking: Research and Experience, vol. 3, no. 3, 1992.
....delays, according to the position of flow in the service hierarchy. Fair queuing (FQ) The fair queuing service discipline was initially proposed by John Nagle [Nagle 85, Nagle 87] and has suffered subsequent modifications and refinements by various authors [Demers 89, Davin 90, Mckenney 90, Keshav 91a, Keshav 91c] This service discipline is based on a conceptual bit by bit round robin (BR) The BR scheme is simulated by calculating the time when a packet would have left the intermediate system using the BR algorithm and inserting that packet into a queue sorted by departure time. In [ZhangHui ....
- Srinivasan Keshav. On the Efficient Implementation of Fair Queueing. Internetworking: Research and Experience, Volume 2, pp. 157-173, 1991.
....above can avoid the cell interleaving problem associated with AAL5, it can be used to support mpt pt connection over the unspecified bit rate (UBR) service for data traffic, where no flow control mechanism is employed. However, more sophisticated packet level weighted fair queueing (WFQ) scheme [8, 10] other than simple roundrobin scheduling will be needed to ensure the fairness of bandwidth allocation among connections with packets of different size. However, WFQ is relatively expensive to implement, and such additional costs defeat the purpose of a low cost UBR service for data traffic. On ....
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, Sept. 1991, vol.2, (no.3):157-73.
....can significantly improve the performance of TCP over UBR service in terms of fairness. Similar results are also reported in [5] C. Fair Queuing Techniques The design of fair queuing techniques for conventional networks has been a subject of intensive research in recent years [3] 7] 8] [10], 13] 15] 18] These works focus on the fair allocation of network resources for different packet streams. The difficulty arises from the fact that the packets are of different sizes. Even if each packet stream is buffered at a separate queue (perconnection queuing) a round robin scheduling ....
S. Keshav. On the efficient implementation of fair queuing. Internetworking: Research and Experience, Sept. 1991, vol.2, (no.3):157-73.
....1. Queueing Mechanism Plus Packet based Scheduling at a Merging Point can be used to support mpt pt connection over the unspecified bit rate (UBR) service for data traffic, where no flow control mechanism is employed. However, more sophisticated packet level weighted fair queueing (WFQ) scheme [7], 9] other than simple round robin scheduling will be needed to ensure the fairness of bandwidth allocation among connections with packets of different size. However, WFQ is relatively expensive to implement, and such additional cost defeats the purpose of a low cost UBR service for data ....
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, Sept. 1991, vol.2, (no.3):157-73.
....Further simulations with di erent randomly chosen start times for the trac generators show an arithmetic mean of about 10 for the variation coecient. As a result, RED seems to be no longer fair when the average queue size oscillates around max th . 5 Trac Phase E ects E ects of trac phases [2] have been one of the reasons to develop queueing disciplines based on random drop. Since the random factor of RED has a small in uence only in case of max p = 0:02 and UDP only trac (cf. Sect. 2.2) the temporal interaction of the CBR trac generators with the average queue size is considered a ....
S. Floyd and V. Jacobson. On trac phase eects in packet-switched gateways. Internetworking: Research and Experience, 3(3):115-156, Sept. 1992.
No context found.
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2(3):157--173, September 1991.
No context found.
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2(3):157-173, September 1991.
No context found.
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2(3):157--173, September 1991.
No context found.
S. Floyd and V. Jacobson, #On Tra#c Phase E#ects in packet Switched Gateways," Internetworking:Research and Experience 3, pp. 115#156, September 1993.
No context found.
S. Keshav. On the efficient implementation of fair queueing. Internetworking: Research and Experience, 2(3):157--173, September 1991.
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