J. F. Shoch and J. A. Hupp. Measured performance of an ethernet local network. Communications of the ACM, 23(12):711--721, 1980.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....models and to offer an example of how to model our measured packet arrival processes, we describe a fitting procedure based on the index of dispersion for counts for the Markov modulated Poisson process. I. INTRODUCTION Since the first analyses of computer traffic in the mid and late 1970s [10, 19], which showed that packet arrival processes are highly variable, researchers have frequently described computer communication patterns as bursty . Yet few have bothered to define burstiness. Most seem to invoke the term bursty when confronted with processes whose interarrival time distributions ....

Shoch, J. F. and J. F. Hupp, Measured Performance of an Ethernet Local Network, Communications of the ACM 23, 12 (December 1980), 711-721.

....frequency components in the data, as seen from Figure 8. The most dominant frequency over the regional segment also corresponds to the work day. However, the other relatively dominant dominant frequencies do not follow the same pattern as the backbone or the cross country segments. Shoch and Hupp [37], and Fowler and Leland [10] have reported a similar strong diurnal pattern in packet arrival rates over a LAN, and over an external campus gateway, respectively. It is interesting to note the similarities in the frequency domain between these studies and wide area end to end delays. Dominant low ....

....has reported existence of similar localities at the level of processes. Feldmeier [7] has used locality of traffic for routing caches in gateways. Gussella [13] has studied local area network traffic and has reported on packet inter arrival and size distributions. Shoch and Hupp s landmark paper [37] is perhaps one of the earliest papers on traffic measurements over a Local Area Network. Traytic Characteristics of the T1 NSFNet BackBone: A recent study by Claffy et al. [4] has reported on the traffic characteristics over the NSFNet backbone. The study gives long term growth of traffic ....

Shoch, J.F., and J.A. Hupp, "Measured performance of an Ethernet local network, "Commun. ACM, pp. 711-721, December 1980.

....represents the short and long packet size, and t denotes the fraction of short packets generated. The bimodal distribution represents the actual situation accurately where the distribution of Ethernet packet size has two peaks with the larger one at the minimum and the smaller one at the maximum [3, 12]; interactive jobs, which take up most of the traffic, generate numerous short packets while large volume messages are carried in the long packets. Fig. 8 shows the throughput delay curves for various values of t. We used p = 64 bytes and p2 = 1518 bytes which respectively correspond to the ....

J. F. Shoch and J. A. Hupp, Measured Performance of an Ethernet Local Network, Communications of the ACM 23 (1980) 711-721. 21

....immediately, the time of the next attempt, after k unsuccessful transmissions, is selected with a uniform distribution on f1; 2; 3; 2 min(10;k) g, until k = 16 when the packet is discarded. Results of measurements of the performance of existing Ethernet installations can be found in [101]. This paper reports results of an experiment that was carried out on 120 stations; very low latency and few collisions were experienced under normal loads, the channel utilization approached 98 under arti cially generated very heavy load. It is because of this remarkable performance that ....

J.F. Shoch, and J.A. Hupp, Measured performance of an Ethernet local network, Comm. ACM 23 (1980) 711-720.

....stationary, that is constant in time in the statistical sense. Naturally in real data this assumption will hold only approximately, and in some cases not at all (for instance see [26] For example diurnal variation in load is a recognised feature of trac in most contexts (for examples see [27] [28], 25] is this also true of H In the seminal paper [1] the authors speculate that this is so, and further that there is a correlation between the load and H . We performed monitoring separately over blocks of data 1, 4 and 24 hours long, with the intention of studying the diurnal and weekly ....

Shoch and Hupp, \Measured performance of an Ethernet local network, " Comm. ACM, vol. 23, no. 12, pp. 711-721, 1980.

....others. Our network load control is not motivated by a need to improve low level network performance or stability. We assume that underlying network resources are stable under heavy load. We evaluate our mechanisms on Ethernet [46] networks, which have been shown to be stable under heavy load [62]. We are instead motivated to control network load for the higher level reasons stated above: to implement administrative policies and use available network resources more effectively to support efficient remote job execution. To implement CPU and network co allocation in the matchmaking ....

J. Shoch and J. Hupp. Measured performance of an ethernet local network. Communications of the ACM, 23(12):711--721, December 1980.

....Unlike centralized applications of the past, distributed multimedia applications run over one or more networks. Network researchers have studied theoretical limits of past and current networks, as well as the practical limits under normal workloads. Networks that have been studied include Ethernet [22], high speed gigabyte networks [8] Asynchronous Transfer Mode (ATM) networks [16] and the Internet. Network researchers have also studied how best to use a network s available bandwidth. Network utilization research includes the effects of delay jitter on application performance and transport ....

John F. Shoch and Jon A. Hupp. Measured performance of an Ethernet local network. Communications of the ACM, 23(12):711--720, December 1980.

....we use Ethernet to mean an unslotted, 1 persistent, carrier sense multiple access method with collision detection and binary exponential backo#. In the past two decades, Ethernet performance has been carefully studied ( 1] 2] 3] 4] 5] 6] 7] 8] 9] 11] 14] 15] 17] 18] 19] 20] 21] [23] [24] 25] 26] 27] 28] 29] Many analytical models have been formulated ( 2] 3] 5] 6] 9] 14] 15] 17] 25] 26] 27] 28] Due to the complexity of the CSMA CD retransmission algorithm and variety of LAN topologies, these analytical approaches employ a number of simplifying ....

....applications, especially when response time is not closely constrained. However, due to the inflexibility of measurement on physical networks, only limited performance measurements under typical network configuration are reported in literature. Other measurement work can be found in [8] [23] [24] By using simulation, performance can be easily measured for various Ethernet topologies and system configurations. Some detailed simulation models are presented in [11] 18] 19] 20] 21] 29] which can be used to model Ethernet with di#erent size, transmission rate, Ethernet length and ....

J. F. Shoch and J. A. Hupp, Measured performance of an Ethernet local network, Communications of ACM, Vol. 23, No. 12, pp. 711-721, December 1980.

....we use Ethernet to mean an unslotted, 1 persistent, carrier sense multiple access method with collision detection and binary exponential backo#. In the past two decades, Ethernet performance has been carefully studied ( 1] 2] 3] 5] 6] 7] 8] 9] 10] 12] 16] 17] 20] 21] 22] 25] 26] [28] [29] 30] 31] 32] 33] 34] Many analytical models have been formulated ( 2] 3] 6] 7] 10] 16] 17] 20] 30] 31] 32] 2 [33] Due to the complexity of the CSMA CD retransmission algorithm and variety of LAN topologies, these analytical approaches employ a number of simplifying ....

....applications, especially when response time is not closely constrained. However, due to the inflexibility of measurement on physical networks, only limited performance measurements under typical network configuration are reported in literature. Other measurement work can be found in [9] [28] [29] By using simulation, performance can be easily measured for various Ethernet topologies and system configurations. Some detailed simulation models are presented in [12] 21] 22] 25] 26] 34] which can be used to model Ethernet with di#erent size, transmission rate, Ethernet length and ....

J. F. Shoch and J. A. Hupp, Measured performance of an Ethernet local network, Communications of ACM, Vol. 23, No. 12, pp. 711-721, December 1980.

....a more aggressive approach to capture the channel is desirable. We will look at two channel access mechanisms, a static and a dynamic adaptive algorithm, and compare their relative merits. Simulation will be used in our study of the integrated service bridge network. 3. SIMULATION MODEL Shoch[3] has shown that the data load on most local networks occupies less than 10 of the network capacity. However, his measurements are more than a decade old and most LANs are populated with machines much faster than those used in his experiment. Hence, in our simulation, 20 data stations are used to ....

Shoch, J F and Hupp, J A "Measured Performance of an Ethernet Local Network," CACM,V ol 23, No 12, Dec 1980, pp 711-721.

....immediately, the time of the next attempt, after k unsuccessful transmissions, is selected with a uniform distribution on f1; 2; 3; 2 min(10;k) g, until k = 16 when the packet is discarded. Results of measurements of the performance of existing Ethernet installations can be found in [93]. This paper reports results of an experiment that was carried out on 120 stations; very low latency and few collisions were experienced under normal loads, the channel utilization approached 98 under artificially gener Randomized Communication in Radio Networks 7 ated very heavy load. It is ....

J.F. Shoch, and J.A. Hupp, Measured performance of an Ethernet local network, Comm. ACM 23 (1980) 711--720.

....factors that determine the speed of data transmission across an internetwork. As network data rates increase, the delay through routers becomes the significant limiting factor to network performance, in a local area context. Intrinsic network delay, although heavily studied [MB76, AL79, TH80, SH80, Gon85, BMK88] is a non issue when routers are added to the system. As we show in this chapter, router delay is, by far, the largest delay component in the Stanford internet. Furthermore, these packet delays have a large variance that cannot be attributed to any cause. Such large variances will ....

....A useful reference is the Poisson density function, which results in a exponential distribution of interarrival times. Analytical models often assume Poisson distributions, providing a baseline for comparison of other results. Unfortunately, real networks do not exhibit Poisson characteristics [SH80, Llo86, PDA86] Llo86] and others have proposed ways to measure burstiness which, unfortunately, require somewhat subjective parameters to define what looks bursty . In section 2.5.4, I propose a simple technique that allows direct comparison to Poisson results. CHAPTER 2. ROUTER DELAY 23 ....

[Article contains additional citation context not shown here]

John F. Shoch and Jon A. Hupp. Measured performance of an ethernet local network. Communications of the ACM, 23(12):711--721, December 1980. Also in CSL-80-2, Xerox Palo Alto Research Center.

.... that has been the isolation of an intermittent fault (bug) found in the gateway software for the Admiral megastream prototype network after three months of testing and analysing performance data [Merabti,86] Another example of performance profiling through measurements is that of Shoch and Hupp [Shoch,80] on the development of the Ethernet local area network. Conducting performance measurements and building the infrastructure for experimentation are complex tasks which could be facilitated by the provision of software hooks when the systems are initially designed and implemented [Svobodova,81] ....

Shoch, J. and Hupp, J., "Measured Performance of an Ethernet Local Network.," communications of the ACM, 23(12), pp. 711720, (December 1980).

....exponentially. The simulations suggest that when n = 2 linear backoff is stable provided only that the sum of the two rates is less than 1. There are other definitions of stability of a transmission protocol. The throughput is the rate at which packets are successfully transmitted. Shoch and Hupp [18] give a throughput definition : The protocol is stable if the throughput is an increasing function of the input rate l. Note the system may achieve this type of stability while packets are piling up or being discarded. Simulations suggest that binary exponential backoff is stable in this sense ....

Shoch, J.F. and Hupp J., "Measured performance of an Ethernet local network", Communications of the ACM , 23, 12 (Dec. 1980), 711-721.

....External traffic (traffic for which one host is not on the LAN) peaked as high as 14.7 of the available Ethernet bandwidth. This 14.7 corresponded to 95 of the available bandwidth on the external T 1 line that connected the LAN to the regional network. Previous studies by other authors (e.g. SH80] had much lower average and peak utilizations. Burstiness analysis shows that the traffic is bursty over a wide range of time scales. There was a striking similarity between the mean packet arrival rate averaged by hour for 307 hours and the same rate averaged by minute for the burstiest 6 ....

J. F. Shoch and J. A. Hupp. Measured Performance of an Ethernet Local Network. Communications of ACM, 23(12):711--721, December 1980. 31

....and uses point to point with acknowledgements to messages to move the broadcast packets around this ring. For a group of N hosts, the protocol requires at least N messages to be sent on the network for each reliable broadcast. The Ethernet is known to have poor performance on small packets [47]. This is largely because the Ethernet channel acquisition protocol enforces a 9.6 sec delay between the end of one packet and the beginning of the next. Since the control packets in the Caching Ring are all small, we expect this to have a serious effect on the performance. The Ethernet limits ....

J. F. Shoch and J. A. Hupp. Measured performance of an Ethernet local network. Communications of the ACM 23(12):711-721, December, 1980.

....Excel on a microcomputer. However, there are two treatments that are both accurate and tractable. Hammond and O Reilly [HaOR86] extend Tobagi and Hunt s analysis [ToHu80] to the 1persistent case and produce a formula for throughput that compares well with the measured performance of Shoch and Hupp [ShHu80] and compares even better with simulation results. Lam [Lam80] provides a formula for delay (mean time in system until successful transmission) that is tractable and compares quite well to the simulations and measurements obtained by Gonsalves and Tobagi [GoTo88] In the NetMod implementation, the ....

J.F. Shoch, J.A. Hupp, "Measured Performance of an Ethernet Local Network," Communications of the ACM, Vol.23, No.12, Dec. 1980, pp.711-721.

....computer network under varying load conditions. A hierarchical approach to traffic characterization provides a deeper insight into system behavior and was adopted for our study. There are several advantages in employing this approach over the traditional techniques for traffic characterization [4, 6, 7, 8, 12]. First, it facilitates the measurement of intra LAN, interLAN, and WAN traffic for each subnet, which are found to demonstrate distinctive patterns. These patterns are important when choosing input parameters or building traffic generators for simulation studies on network and protocol ....

....often made that may give results which do not conform too well to the real situation. The success of analytical and simulation techniques lies in the careful and accurate choice of input parameters and representation of the network. Real time measurements can be used to identify these parameters [1, 2, 3, 12] and was therefore used in our study. Realtime measurements help model real life situations accurately and can be used to improve network models iteratively. Real time measurements provide valuable insights into network usage and behavior and facilitate the performance evaluation of operational ....

Shoch, J. F. and Hupp, J. A., Measured performance of an ethernet local network, Communications of the ACM, Vol. 23, No. 23, December 1980.

....Internet and Intranet networks. Our goal is to analyze the ability of the existing LAN segments to carry voice traffic. Ethernet is one of the most common LAN technologies in the market today and many capacity measurement and performance studies have been done to determine its capabilities. [4,5,6,7] Some of the original assumptions and conditions which eliminated Ethernet from consideration for voice traffic have changed, and have made it worth another look. Ethernet s CSMA CD utilizes the Binary Exponential Backoff algorithm for contention resolution. Studies have shown this scheme causes a ....

J. Shoch and J. A. Hupp, "Measured Performance of an Ethernet Local Network," Communications of the ACM, vol. 23, no. 12, December 1980, pp. 711-721.

....are conservative estimates for blast interruptions at a file server because we do not use any caching on the workstations. This significantly increases the traffic to the file server, and hence the number of interrupted blasts. 7 Related Work Locality in network traffic has been noted earlier [3, 8, 10]. However, to the best of our knowledge, no protocol implementation has taken advantage of this phenomenon to the extent described here. In his efforts to improve TCP performance, Jacobson predicts that the next incoming TCP packet on a given connection will be the next packet on that connection ....

J.F. Shoch and J.A. Hupp. Measured performance of an Ethernet local network. Communications of the ACM, 23(12):711--721, December 1980.

....for packet setup and teardown and the time spent for protocol conversion by the gateways has to be considered. Once defined the set of parameters characterizing the workload of a network, the appropriate measurement tools must be found. In most of the studies presented in the literature (e.g. [27], 28] 29] two different approaches are adopted: either special purpose devices are used or adhoc test and monitoring tools are constructed. In what follows, we give a survey of a few studies appeared in the literature which describe how to measure and analyze the parameters above introduced ....

....In what follows, we give a survey of a few studies appeared in the literature which describe how to measure and analyze the parameters above introduced and which of them are to be used for the input definition of both analytical and simulation models of different types of networks. In [27] and [28] the authors focus on experimental measurements of Ethernet local area networks. As we will see, although the two studies are quite similar, the results obtained by Shoch and Hupp are different from Gusella s because of the architectural differences of the two environments. In [27] the ....

[Article contains additional citation context not shown here]

J.F. Shoch and J.A. Hupp. Measured Performance of an Ethernet Local Network. Comm. of the ACM, 23(12):711--721, 1980.

....the importance of realistic traffic characterization with an example drawn from a simulation study of the packet loss and delay effects induced by one LAN to BISDN access proposal. Although there have been many studies of LAN traffic since the early Ethernet measurements of Shoch and Hupp, [1] the emphasis has been on intermediate time scale behavior [2] 3] 4] or on user oriented measures of behavior such as LAN throughput and delay [5] 6] More seriously for our study, little published data are available on the packet loss and timestamp errors introduced by the data ....

....Measurement 5 3.1 Arrival Rates The most familiar characterizations of packet traffic are arrival rates over medium to long time scales: the number of bytes or packets seen per second, hour, or day. The classic 1979 study of a 2. 94 megabit per second Ethernet by Shoch and Hupp [1] reported seeing about 2.2 million packets per day, representing less than 1 utilization of the network capacity, with a pronounced diurnal cycle. The diurnal cycle is readily apparent in our current measurements of aggregate traffic, both for internal and external traffic (as illustrated for ....

Shoch, J. F. and J. A. Hupp, "Measured Performance of an Ethernet Local Network", pp. 711-721, Communications of the ACM, Volume 23, Number 12 (December 1980)

....difficult [2] Most LAN topologies are based on sharing the bandwidth fairly and hence do not provide guaranteed quality of service for realtime multimedia applications in general. Since some past studies on LAN traffic analysis have reported that LAN traffic shows high variability [1] 4] 7] [11], the bandwidth sharing property may result in prevalent delay uncertainty or packet loss, which may be a serious limitation for the successful design of LAN based interactive multimedia applications. The Xphone system was developed to test the feasibility of LAN based interactive multimedia ....

J. F. Shoch and J. A. Hupp. Measured performance of an Ethernet local network. Commun. ACM, 23(12):711--721, December 1980.

No context found.

J. F. Shoch and J. A. Hupp. Measured performance of an ethernet local network. Communications of the ACM, 23(12):711--721, 1980.

No context found.

J. F. Shoch and J. A. Hupp, "Measured Performance of an Ethernet Local Network," Communications of the ACM, Vol. 23, no. 12, December 1980.

First 50 documents

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