L. N. Bhuyan, D. Ghosal, and Q. Yang, "Approximate analysis of single and multiple ring networks," IEEE Transactions on Computers, vol. 38, no. 7, pp. 1027--1040, July 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....[4] for these three, the multiple token approach is shown to offer better delay performance than the other two. In an attempt to reduce the waiting for a free token and to further remedy the problems of low utilization and higher queueing delays at high data rates, protocols with multiple rings [3] and multiple tokens [7] are also proposed. The potential impact on the performance of them are that when multiple channels are used, messages are transmitted on each of the channels independently so that there is an increased bandwidth, there exist multiple paths between a source destination pair ....

....is an increased bandwidth, there exist multiple paths between a source destination pair giving improved fault tolerance and due to multiple tokens the time between consecutive network accesses and in turn the token interarrival times are reduced leading to decrease in wait time for the packets. In [3], the authors proposed an approach of multiple rings hav ing a token on each ring. An approximate analysis of their model shows that delay performance is better than a single ring with single token, analysed in [12] But as the number of rings increases it is likely that all the tokens move ....

[Article contains additional citation context not shown here]

L. N. Bhuyan, D. Ghosal, and Q. Yang. Approximate analysis of single and multiple ring networks. IEEE Trans.Comput. , C-38(7):1027--1040, 1989.

....that the miss frequency is different from the 28 miss rate MR which is the probability of a cache miss per reference; the period during which the processor is blocked is included in MF . To calculate the transmission delay T on the slotted ring, the approximated M=G=1 model by Bhuyan et al. in [6] is used with a slight modification to suit to our problem. The transmission delay T consists of W and x, where W is the mean waiting time for a message in the transmission queue and x is the time to actually transmit the message packets. Here, the simplest case where the messages consist of a ....

....queue (Q(d x) where Q is the number of messages ahead in the transmission queue, and is equal to W ) Now, d = U= 1 Gamma U) where U is the ring utilization. It is assumed that successive slots behave independently as the error produced by this approximation is known to be relatively small [6]. U is given by ECDx where is the average message transmission rate. Since only the non local misses transmit messages on the ring, 1 Gamma P L )MF . The time to access the main memory M is t m plus the time to complete the remaining accesses in the memory queue. Similar to the calculation ....

[Article contains additional citation context not shown here]

L. Bhuyan, D. Ghosal and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks," Transactions on Computers, IEEE, Vol. 38, No. 7, 1027-- 1040, July 1989.

....as a closed model. In a closed system, if we can identify and distinguish the effects among multiple servers, it is still possible to use an open model to reasonably capture the performance characteristics of the system. Examples of using an open model to study a closed ring system are reported in [5] for network analysis and in [4] for cache coherence protocols. The modeling results in [4] are well supported by the simulation. Besides its practical applications in performance analysis, there are three reasons why using an open model to study the ring hot spot effects is acceptable and ....

.... Gamma l ) 2.7) Because the successive slots have been assumed to behave independently, the probability of a slot on a cool local ring to be full is U l . The time needed for a packet in each station to find an empty slot can be approximated to have a geometric distribution. The analyses in [5] show that the error caused by the independence assumption is trivial. The probability of this event is U i l (1 Gamma U l ) Thus, the average time of finding an empty slot is d n = 1 X i=0 it r (1 Gamma U l )U i l = t r U l 1 Gamma U l = nt 2 r (1 h 2(1 Gamma h ) 1 ....

L. N. Bhuyan, D. Ghosal and Q. Yang, "Approximate analysis of single and multiple ring networks ", IEEE Transactions on Computers, Vol. 38, No. 7, July 1989, pp. 1027-1040.

.... this method lies between that of detailed simulation of the SCI communication protocol, where the simulator keeps track of every symbol travelling on the ring [8, 7, 24] and, that of trace postprocessing [23] or statistical analysis, where the network simulator is driven by random access patterns [6]. 4 Simulation Results The results of the experiments are presented in this section. First, the workload is presented and the behaviour of the programs discussed. Then, the following are examined in turn: influence of coherent cache size and latency, bandwidth and round trip delay and, ....

L N Bhuyan, D Ghosal, and Q Yang. Approximate analysis of single and multiple ring networks. IEEE Trans. on Computers, C-38(7):1027--1040, July 1987.

....the message to TC(i 1) if the bus is not busy or store the data in the FIFO if the bus is busy. Block i F I F O F I F O FIFO From Block i 1 From Block i 1 Figure 3: The TC in a register insertion ring A comparison of these three rings with similar routing protocol is given in [20]. The token ring is always the worst one. Slotted and register insertion rings have similar results, even if the register insertion rings shows better results for small flows. As opposed to the slotted ring, only the busses where a transfer is occurring are active in the register insertion ring. ....

L. N. Bhuyan, D. Ghosal, Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", in IEEE Transactions on Computers, vol. 38, no. 7, pp. 1027-1040, July 89.

....the numbers of header and data message transmissions. This is because on the crossbar switch network, a message only takes one hop to reach any destination PE. These utilization parameters are combined with the M G 1 queue model to estimate the execution time of applications on both architectures [3]. Thus, ET itself is a function of ET . ET = Refs L1Miss tL2 L2Miss L2MissLatency(ET ) The derivation of ET is iterated until it converges to a tolerant level ( 0.1 ) We use two metrics to evaluate the performance of two architectures. One is the ratio of execution times, namely, the ....

L. Bhuyan, D. Ghosal and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", Transactions on Computers, IEEE, Vol. 38, No. 7, 1027--1040, July 1989.

....1=MR on the average, and then it stalls for the period L due to a cache miss. Next, we break down L in to four components: The transmission delay T is the time for a PE to transmit a message (read request, data, or invalidation) To calculate T , we use approximated M=G=1 model by Bhuyan et al. in [6] with slight modification to suit to our problem. The propagation delay P is the time for a message to reach the destination PE and is determined by number of hops. P is obtained from our model of communication locality and will be described below. The memory access latency M is the memory access ....

L. Bhuyan, D. Ghosal and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", Transactions on Computers, IEEE, Vol. 38, No. 7, 1027--1040, July 1989.

....of new multicast protocols such as IP multicast[6] A drawback of the shared medium is that it typically permits only a single terminal to transmit at any time, thereby not utilizing all network segments efficiently. A scheme allowing the capacity of the medium to be reused may be designed [7, 8, 9, 10, 11], but this is often at the cost of complexity in the high speed access control hardware. Access control mechanisms for a shared medium also depend strongly on the size of the network and are usually efficient only for local area environments. The two main types of network are connection oriented ....

....transmissions in the same slot over disjoint segments of the bus. Slot reuse is a general method to better utilize shared links in ring and bus networks. The slot reuse algorithms in DQDB [11, 7] Simple [19] and CRMA [20] depend on control information in the slots. Buffer insertion networks [10, 8] when combined with destination release as in METARING [9] may reuse capacity of individual links and resolve eventual contention by delaying the incoming packet stream through an elastic buffer. With slot reuse, the access scheme complexity is increased, regardless of whether it is done in ....

L. Bhuyan, D. Ghosal, and Q. Yang, "Approximate analysis of single and multiple ring networks," IEEE Transactions on Computers, vol. C-38, no. C-38, pp. 1027-- 1040, 1989.

....extensive analysis in the context of Local Area Networks (LANs) since the mid 70 s. Hafner et al. [18] proposed a register insertion ring LAN in 1974. Hooper and Needham describe the architecture of the Cambridge Fast Ring [20] that uses the slotted ring approach in a LAN environment. Bhuyan et al. [5] among several others have modeled the performance of ring LANs. Still in the context of distributed systems, Delp et al. proposed a token ring distributed shared memory system (Memnet [9] with cache coherence maintained in hardware by means of a snooping like coherence protocol. More recently ....

L. Bhuyan, D. Ghosal, and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", IEEE Transactions on Computers, Vo. 38, No. 7, pp. 1027-1040, July 1989.

....the mix of slots matches the expected mix of messages. The particular mix depends on the cache coherence protocol, and therefore we postpone this discussion until the next Section. A slotted ring can be seen as a generalization of a token ring in which slots represent the tokens. Bhuyan et al. [4] use an analytical framework to compare the performance of slotted and register insertion rings. Their overall conclusion is that under light loads the register insertion ring may exhibit faster access times than a slotted ring since a message does not wait for a proper slot to pass by. On the ....

....it is important to determine the rates of probes and block messages. These rates can be obtained from Equations 2 and 3 as follows where N proc is the number of processors in the system. We now derive a simple performance model of the slotted ring similar to the one used by Bhuyan and others in [4]. The throughput of the slotted ring for probes and block messages is as follows: TABLE 5. Parameters from trace driven simulations of the program Parameter Definition N cyc Total number of instructions executed in a given processor N lmiss Total number of misses to local memory by a processor N ....

[Article contains additional citation context not shown here]

L. Bhuyan, D. Ghosal, and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", IEEE Transactions on Computers, Vo. 38, No. 7, pp. 1027-1040, July 1989.

....of transactions or independent sequential programs. We restrict our attention to the ring, which is an important network in both theory and practice. From a theoretical perspective, the ring is a basic network structure, and much work has been done on developing and analyzing algorithms for it [3, 5, 9, 15, 16, 20, 21]. In practice the ring is either the basis of or an essential component of many parallel and distributed architectures [12, 13, 22, 23, 19] The element of our model which most differentiates it from previous scheduling algorithms is that task migration from one processor to another takes time ....

L. Bhuyan, D. Ghosal, and Q. Yang. Approximate analysis of single and multiple ring networks. IEEE Transactions on Computers, 38:1027--1040, 1989.

....of the protocol in hard real time environments. On the other hand, many studies of CSMA CD and token ring protocols for distributed hard real time applications have been conducted. The issues in design and analysis of deadline driven communication protocols for CSMA CD networks are addressed in [5, 29, 32, 42, 50, 57, 65, 66]. The real time performance of various token ring protocols are considered in [30, 37, 50, 55, 63] Our work reported in this paper complements the previous studies by addressing the issues pertinent to hard real time communication in a high speed network where the timed token medium access ....

L. N. Bhuyan, D. Ghosal and Q. Yang, "Approximate analysis of single and multiple ring networks," IEEE Transactions on Computers, Vol. 38, No. 7, pp 1027-1040, July 1989.

....numbers of header and data message transmissions. This is because on the crossbar switch network, a message only takes one hop to reach any destination PE. These utilization parameters are combined with the M G 1 queue model to estimate the execution time of applications on both architectures (see [11] for the details of T on ring networks) Thus, ET itself is a function of ET . ET = Refs Miss MissLatency(ET ) The derivation of ET is iterated until it converges to a tolerant level ( 0.1 ) We use two metrics to evaluate the performance of two architectures. One is the ratio of execution ....

L. Bhuyan, D. Ghosal and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", Transactions on Computers, IEEE, Vol. 38, No. 7, 1027--1040, July 1989.

....the numbers of header and data message transmissions. This is because on the crossbar switch network, a message only takes one hop to reach any destination PE. These utilization parameters are combined with the M G 1 queue model to estimate the execution time of applications on both architectures [11]. Thus, ET itself is a function of ET . ET = Refs L1Miss tL2 L2Miss L2MissLatency(ET ) The derivation of ET is iterated until it converges to a tolerant level ( 0.1 ) We use two metrics to evaluate the performance of two architectures. One is the ratio of execution times, namely, the ....

L. Bhuyan, D. Ghosal and Q. Yang, "Approximate Analysis of Single and Multiple Ring Networks", Transactions on Computers, IEEE, Vol. 38, No. 7, 1027--1040, July 1989.

....For a V station, the probability of arrival of a full slot is the same as the utilization, U v , of a slot by the voice traffic. U v can be approximated by the product of aggregate voice packet arrival rate as seen by a slot and the time required for a slot to flow from one station to the next one [2, 18] (which is simply (x R=N ) assuming stations are uniformly distributed on the ring) The aggregate voice 8 This implies that a slot takes x time units to pass a station. packet arrival process seen by a slot is the superposition of N independent MMPP sources. It can be shown that the ....

L. N. Bhuyan, D. Ghosal, and Q. Yang. Approximate Analysis of Single and Multiple Ring Networks. IEEE Transactions on Computers, 38(7), July 1989.

....that the miss frequency is different from the miss rate MR which is the number of memory accesses per reference the period during which the processor is blocked is included in MF . To calculate the transmission delay T on the slotted ring, we use approximated M=G=1 model by Bhuyan et al. in [8] with slight modification to suit to our problem. The transmission delay T consists of W and x, where W is the mean waiting time for a message in the transmission queue and x is the number of packets in the message. We explain the simplest case: the transmission delay of a header message, that ....

....in the queue (Q(d x) where Q is the number of messages ahead in the transmission queue, and is equal to W ) Now, d = U= 1 Gamma U ) where U is the ring utilization. We assume that successive slots behave independently as the error produced by this approximation is known to be relatively small [8]. U is given by ECDx where is the average message transmission rate. Since only the non local misses transmit messages on the ring, 1 Gamma PL )MF . The time to access the main memory M is t m plus the time to complete the remaining accesses in the memory queue. Similar to the calculation ....

L. Bhuyan, D. Ghosal and Q. Yang, Approximate Analysis of Single and Multiple Ring Networks, Transactions on Computers, IEEE, Vol. 38, No. 7, 1027--1040, July 1989.

....a good reference is the textbook authored by Leighton [22] We focus on the ring architecture, which is an important network in both theory and practice. From a theoretical perspective, the ring is a basic network structure, and much work has been done on developing and analyzing algorithms for it [5, 7, 16, 22, 24, 30, 31]. In practice the ring is either the basis of or an essential component of many parallel and distributed architectures [20, 28, 32, 33] Our primary vehicle for the evaluation of an algorithm is by a worst case analysis of the quality of the schedule it produces. Our goal will be to schedule all ....

L. Bhuyan, D. Ghosal, and Q. Yang. Approximate analysis of single and multiple ring networks. IEEE Transactions on Computers, 38:1027--1040, 1989.

....of transactions or independent sequential programs. We restrict our attention to the ring, which is an important network in both theory and practice. From a theoretical perspective, the ring is a basic network structure, and much work has been done on developing and analyzing algorithms for it [4, 6, 10, 16, 17, 21, 22]. In practice the ring is either the basis of or an essential component of many parallel and distributed architectures [14, 20, 23, 24] The element of our model which most differentiates it from previous scheduling models is that task migration from one processor to another takes time ....

L. Bhuyan, D. Ghosal, and Q. Yang. Approximate analysis of single and multiple ring networks. IEEE Transactions on Computers, 38:1027--1040, 1989.

....to slotted rings used in some earlier proposed methods [8, 9, 10] Most of the traffic in cache coherence environment is bursty, instead of uniformly distributed over time. Under such load conditions (i.e. low average load) an insertion ring would definitely perform better than slotted ring [17]. Exhaustive simulations have shown that for different types of sharing patterns, from 19 60 70 80 90 110 Channel 8 16 64 Width (in bits) Time ( of full map) MP3D: 8000 Molecules, 64 Processors 4 Sub Rings, 2 Pointers Figure 13: Simulation results for MP3D for different ....

L. N. Bhuyan, D. Ghosal, and Q. Yang, "Approximate analysis of single and multiple ring networks," IEEE Transactions on Computers, vol. 38, no. 7, pp. 1027--1040, July 1989.

No context found.

L. Bhuyan, D. Ghosal, and Q. Yang, "Approximate analysis of single and multiple ring networks, " IEEE Tr. on Computers, vol. 38, pp. 1027--1040, Jul. 1989.

No context found.

L. N. Bhuyan, D. Ghosal and Q. Yang, "Approximate analysis of single and multiple ring networks", IEEE Trans. Computers, Vol. 38, No. 7, 1989, pp. 1027-1040.

No context found.

L. N. Bhuyan, D. Ghosal and Q. Yang, "Approximate analysis of single and multiple ring networks," IEEE Transactions on Computers, Vol. 38, No. 7, pp 1027-1040, July 1989.

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