S.-K. Kweon and K. G. Shin. Traffic controlled rate monotonic priority scheduling. In Proceedings of IEEE Infocorn 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... scheme which is able to provide end4o end statistical QoS in a network of Generalized Processor Sharing (GPS) schedulers is developed in [14] End to end statistical QoS guarantees are also provided by the scheme proposed in [25] which employs Traffic Controlled Rate Monotonic Priority Scheduling [24]. Our approach was developed independently of [14] 25] and was first presented in [41] 42] In this article we extend our approach and present it in a comprehensive manner. The GPS based scheme [14] is further refined in [22] Schemes for providing end4o end statistical QoS in a network of ....

S.-K. Kweon and K. G. Shin. Traffic controlled rate monotonic priority scheduling. In Proceedings of IEEE Infocorn 1996.

....of research done at the ATM network level. Research at this level is concerned with traffic specification models, CAC (Connection Admission Control) algorithms, and cell scheduling algorithms that can provide deterministic or statistical real time guarantees for video, audio and data traffic types [5, 6, 8, 9, 10, 15, 18, 21, 20, 25]. Network level research also addresses new protocols and service models, such as the work presented in [7, 22, 24] Research at the middleware level has just begun. Here research is needed to close the gap between what realtime and multimedia application designer needs and what the ATM network ....

....in research and what the actual ATM network will provide. We then show how real time communication can be mapped onto ATM CBR and rt VBR service categories. We demonstrate that real time periodic communication is not necessarily best implemented with CBR, although some of the previous research [9, 15] has in essense assumed the CBR traffic model as the hard real time communication traffic model. We then show the limitations of ATM services and how User Level Multiplexing can be used to overcome the limitations. 1 . The rest of the paper is organized as follows. In Section 2, we describe the ....

[Article contains additional citation context not shown here]

S.-K. Kweon and K. G. Shin. Traffic-Controlled RateMonotonic Priority Scheduling of ATM Cells. to appear in Proceedings of the 15th IEEE INFOCOM, March 1996.

....where traffic from different connections is multiplexed. Drawing on the similarities with CPU scheduling, classical real time scheduling policies such as the First Come First Serve (FCFS) Earliest Deadline First (EDF) Generalized Processor Sharing (GPS) Fair Queuing (FQ) etc. are employed [2, 12, 13, 16, 23, 24, 25]. However, most commercially available switches use a high priority queue for HRT connections, and this queue is served in a FCFS manner. In this paper, we focus on the third approach. That is to control the network delays by regulating the input traffic of each connection. Regulating the input ....

....increases with a proper choice of leaky bucket parameters at the UNI. While we focus on traffic regulation for meeting end to end deadlines, our work also complements much of the previous studies which essentially concentrate on designing and analyzing scheduling policies for ATM switches [2, 5, 6, 8, 10, 11, 12, 13, 16, 20, 21, 23, 24, 25]. A modified FCFS scheduling scheme called F IFO was proposed and studied in [2] The switch scheduling policy called Stop and Go is presented in [8] A virtual clock scheduling scheme in which cells are prioritized by a virtual time stamp assigned to them, is discussed in [25] The use of ....

[Article contains additional citation context not shown here]

S. Kweon and K. G. Shin. Traffic-controlled rate-monotonic priority scheduling scheduling of atm cells. In Proceedings of the IEEE Infocom'96, pages 655--662, Mar. 1996.

....flexibility reasons. Had one traffic controller been assigned to each real time channel, a traffic controller must be assigned (de assigned) every time a realtime channel is established (terminated) Although the traffic controller can be implemented with a modified calendar queue as discussed in [23], its scalability is limited for two reasons. First, the number of the time bins of the calendar queue depends on the range of logical arrival times. For instance, if the link capacity is 1 Gbps and if the minimum throughput that can be allocated to a real time channel is 1 Kbps, a cell of a ....

S. Kweon and K. G. Shin, "Traffic-controlled rate-monotonic priority scheduling," in Proc. of IEEE INFOCOM, pp. 655--662, 1996.

....this paper, we change the TCRM s buffer size from 1 to N in order to reduce the cell loss probability when multiple cells arrive at the macro channel in a very short time. The cell drain rate from the buffer is ensured to be ae=L using the cell logical arrival times. Since the admission control in [44] requires only ae as a parameter, the change of the buffer size does not require any other modifications to the structure of the TCRM. Within a macro channel, we do not differentiate statistical real time channels from one another. All the cells arriving at this macro channel are transmitted on a ....

....of the queueing system as illustrated in Figure 4. Although there can be multiple queueing systems when multiple macro channels run through a link, we specify only a single macrochannel since different macro channels are virtually isolated from one another, thanks to the TCRM s Firewall property [44]. When the cell arrival rate is in an M=D=1=N system, the cell blocking probability P b ( can be calculated using the following O(N ) algorithm [45] 0 = 1; 3.1) k 1 = a 0 (OE k Gamma j a k Gammaj 1 Gamma a k ) 0 k N Gamma 1; 3.2) OE 0;n = n k=0 k ) 1 P ....

S.-K. Kweon and K. G. Shin, "Traffic-controlled rate-monotonic priority scheduling," in Proc. of IEEE INFOCOM, pp. 655--662, 1996.

....to be ae i during any time interval of length L=ae i , implying that TCRM emulates circuit switching in the cell level. This feature allows us to provide the CBR services in addition to real time communication services without losing the statistical multiplexing gain of ATM networks. In [18], a simple schedulability test algorithm is derived based on Eq. 2.3) 2.4 Bounding End to End Delays in Multi hop Connections Using the fact that TCRM guarantees the minimum throughput ae i for channel i, we can derive the end toend delay bound of channel i. Given the input traffic ....

S. Kweon and K. G. Shin, "Traffic-controlled ratemonotonic priority scheduling", Real-Time Computing Laboratory Technical Report, May 1995.

....this paper, we change the TCRM s buffer size from 1 to N in order to reduce the cell loss probability when multiple cells arrive at the macrochannel in a very short time. The cell drain rate from the buffer is ensured to be p L using the cell logical arrival times. Since the admission control in [48] requires only p as a parameter, the change of the buffer size does not require any other modifications to the structure of the TCRM. Within a macrochannel, we do not differentiate statistical real time channels from one another. All the cells arriving at this macrochannel are transmitted on a ....

....of the queueing system as illustrated in Fig. 4. Although there can be multiple queueing systems when multiple macrochannels run through a link, we specify only a single macrochannel since different macrochannels are virtually isolated from one another, thanks to the TCRM s Firewall property [48]. When the cell arrival rate is A in an M D 1 N system, the cell blocking probability bOO can be calculated using the following O(N 2) algorithm [49] 1, 3.1) k a 1 ( ljak j l ak) k l 0 k j=l 0 k N 1, 3.2) n 1 50, 5t) n = 0, N (3.3) k=0 1 Pb, X) 1 0, v ....

S.-K. Kweon and K.G. Shin, "Traffic-Controlled Rate-Monotonic Priority Scheduling," Proc. IEEE INFOCOM, pp. 655-662, 1996.

....model [14, 15] Although their work is theoretically attractive, it is not clear whether it can be applied to real life systems since their scheme assumes an infinite buffer at each node. In addition, the implementation complexity of PGPS must be resolved for use in high speed networks like ATM [51]. Effective bandwidth has been investigated in order to provide statistically guaranteed QoS in ATM networks [18, 29, 39] This approach is based on the large deviation theory and often employs an on off process as a source traffic model. In particular, Elwalid et al. 18] derived the worst cast ....

....(2) is simple enough to operate in a high speed switching environment like ATM networks; and (3) requires only a very small buffer space for each real time channel. ffl A framework to provide statistical real time communication services for VBR video data in point to point networks based on TCRM [51]. Since TCRM was originally developed for providing deterministic real time communication services and does not allow statistical multiplexing among real time connections, we make slight modifications to TCRM so that it may allow statistical multiplexing among a set of real time connections. By ....

[Article contains additional citation context not shown here]

S. Kweon and K. G. Shin, "Traffic-controlled rate-monotonic priority scheduling," in Proc. of IEEE INFOCOM, pp. 655--662, 1996.

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