R. Gusella. A measurement study of diskless workstation traffic on an Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, Sept. 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....approximating the average network delay. We consider delays due to LANs and bridges. The topology is modelled as a network of queues. Different measurement and performance studies have shown that LAN traffic is bursty and consists of batches or trains of packets because of the existing protocols [18,19]. In order to account for the burstiness of the traffic and to have a computationally inexpensive measure, we use MX M 1 queues with batch Poisson arrivals. Although it is simple, MX M 1 queueing model may achieve an acceptable fit for the busiest periods of the network [19] and give us the ....

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet," [EEE Tr. on Communications, pp. 1557-1568, September 1990.

....8 10 6 1 10 7 Number of packets Time Unit Time [sec] Time Unit = 10000 slots Figure 4: Behavior of the Markov process over 4 timescales, a=10 q=0. 9 number of states=5 11 5 Fitting Based on measurements, a lot of fitting procedures have been proposed in the litterature (see for example [14], 15] Ours is based on the Markov chain described in (section 2) Here, we fit only two parameters: mean and Hurst parameter (plus the number of states in the Markov chain) As seen in (section ) EX equals (1 Gamma (1=q) 1 Gamma (1=q) For a given EX , it is quite easy to find q ....

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet," IEEE Transactions on Communications, vol. 38, September 1990.

....of computer networks, scientists have studied the size, frequency, and distributions of both for network traffic. Such studies consistently show that the majority of traffic (by packet count) consists of short messages. This property is remarkably sta ble across networks, time, and applications [11, 16]. In a study of traffic on a Ethernet connecting diskless workstations to file servers [11] Gusella found that the majority of packets were less than 576 bytes; of these 60 were 50 bytes or less. In another study [16] Kay et al. measured the TCP and UDP traffic on a FDDI LAN of Unix ....

....for network traffic. Such studies consistently show that the majority of traffic (by packet count) consists of short messages. This property is remarkably sta ble across networks, time, and applications [11, 16] In a study of traffic on a Ethernet connecting diskless workstations to file servers [11], Gusella found that the majority of packets were less than 576 bytes; of these 60 were 50 bytes or less. In another study [16] Kay et al. measured the TCP and UDP traffic on a FDDI LAN of Unix workstations in a university computer science department. They found that TCP message sizes are small: ....

R. Gusella. A measurement study of diskless worksta- tion traffic on Ethernet. IEEE Transactions on Com- munications, 38(9):1557-1568, September 1990.

....performed already. Investigations into the overall capacity of networks investigated, such as that presented in Boggs et al. 17] is important when the efficiency of the computer network can play a significant role in the performance of an attached machine. Additionally, the work of Gusella [39] into the usage of these networks can assist in the planning for future networks. Gusella recorded the header information of every Ethernet network transaction (packet) over a given time period and later processed that data off line. In this way, a partial trace of the operations on the network ....

....data to reduce the quantity of information ultimately recorded. Blaze, who constructed the rpcspy system of reference [11, 12] has used it to good effect for the monitoring of a network based around a large file server. His research results have also been used in several other publications [39, 10, 15, 16, 13]. Dahlin et al. 24] used the rpcspy tools to characterise file system load in a distributed system. The results of this research have then been used to justify the building of a new style of distributed file system, xFS [120] Anderson [1] used rpcspy to analyse the distribution of traffic across ....

Gusella, R. A Measurement Study of Diskless Workstation Traffic on Ethernet. IEEE Transactions on Communications 38, 9 (September 1990), 1557--1568.

.... analytic technique [NEU89] In LAN MAN network performance evaluation studies, the data traffic arrival process is often modelled as a Poisson process or Batch Poisson process, though a number of traffic studies have shown that packet interarrivals are not exponentially distributed ( CIN94] [GUS90], GUS91] LEL94] PAX94a] PAX94b] that packets are structured into batch of cells, and that successive batches are correlated. We will show below that there is experimental evidence that the Batch Markov Modulated Poisson Process (BMMPP) is a good approximation for the aggregate cell ....

....(see Figure 3) 4. Arrival Process Characterization Generally, when modelling data traffic, packet arrivals are often assumed to be Poisson processes. However, a number of studies have shown that the distribution of packet interarrival times significantly differs from the exponential distribution [GUS90], GUS91] PAX94b] LEL94] CIN94] Specifically, in our analysis we represent the packet interarrival 8 times at station N by means of a 2 state MMPP process 1 . Unlike renewal models, MMPP can represent correlation between interarrival times. When the Markov chain is in state i i ....

[Article contains additional citation context not shown here]

R. Gusella, "Measurement Study of Diskless Workstation Traffic on an Ethernet", IEEE Transactions on Communications, Vol. COM-38, pp. 1557-1568, Sept. 1990.

....Demand assigned multiple access works well for bursty sources only if the sizes of the data packets far exceed the size of the DAMA request packets. If this is not the case, then DAMA performs no better than random access. Studies have shown a strongly bi modal distribution of TCP packet sizes (Gusella 1990)(Caceres et al. 1991) due mainly to the fact that most TCP packets are either data carrying packets, which are close to the maximum transmit unit (MTU) size, or acknowledgement packets, which tend to be less that 100 bytes. Since the uplink channel will be used by the HTTP client, and since the ....

Gusella, R. (1990). A measurement study of diskless workstation traffic on an Ethernet. IEEE Trans. Comm. 38, 1557-- 1568.

....theory, based on Poisson models that emphasize at most fleeting correlations between packet arrivals, one can generally engineer a packet switched network to have as low a packet loss rate as desired. Operational experience, however, has been quite contrary and brutal to the Poisson framework [JR86, G90, FL91, DJCME92, PF95], which appears woefully inadequate for accurately predicting actual network behavior. Recent years have seen the rise of self similar traffic models, in which correlations are extremely long lived and have a fractal structure, leading to burstiness on all time scales [LTWW94] Fractal models ....

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet," IEEE Transactions on Communications, 38(9), pp. 1557-1568, September 1990.

....well established principles of TCP IP network workloads: network workloads are bimodal, LAN workloads are dominated by small NFS packets and a small fraction of large NFS packets, and nearly all packets on WANs are less than 512 bytes long. These points are made convincingly for LANs in BibRef[29] (keeping in mind that 1k ND packets have been replaced with 8k NFS packets) and for WANs by BibRef[12] 26 2.7 Conclusions In summary, although, as this chapter has shown, most aspects of network software performance have been investigated to some degree, issues in reducing minimal software ....

....the San Diego Supercomputer Center. This is the same trace as was described and analyzed in BibRef[18] it was taken between 2 and 3 PM on March 23, 1993. Both traces show strong bimodal behavior. This and the other traffic behaviors observed conform to findings of earlier studies BibRef[12]BibRef[29]BibRef[47] Packet size is defined to mean the amount of actual user data sent, not including protocol headers. Figures 1a b show the packet size distributions for TCP and UDP packets. 3.2.1 LAN Traffic The trace that reflects widest usage is a trace of traffic destined for and generated by a ....

[Article contains additional citation context not shown here]

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet," IEEE Transactions on Communications, 38(9), pp. 1557-1568, September 1990. 91

....of the process. Here we introduce the domain of validity which is the domain where H l 6= 0:5, which is very appearent for the 3 and 5 state processes in Figures 2, 3 and 4. 4 Fitting Based on measurements, a lot of fitting procedures have been proposed in the literature (see for example [16], 17] Ours is based on the Markov chain described in (section 2.2) Here, we fit only two parameters: mean and Hurst parameter (plus the number of states in the Markov chain) As seen in (section 2.4) E[X] equals (1 Gamma (1=q) 1 Gamma (1=q) n ) For a given E[X] it is quite easy to ....

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet," IEEE Transactions on Communications, vol. 38, September 1990.

....use of computer networks, scientists have studied the size, frequency, and distributions of both for network traffic. Such studies consistently show that the majority of traffic (by packet count) consists of short messages. This property is remarkably stable across networks, time, and applications [13,18]. In a study of traffic on a Ethernet connecting diskless workstations to file servers [13] Gusella found that the majority of packets were less than 576 bytes; of these 60 were 50 bytes or less. In another study [18] Kay et al. measured the TCP and UDP traffic on a FDDI LAN of Unix ....

....for network traffic. Such studies consistently show that the majority of traffic (by packet count) consists of short messages. This property is remarkably stable across networks, time, and applications [13,18] In a study of traffic on a Ethernet connecting diskless workstations to file servers [13], Gusella found that the majority of packets were less than 576 bytes; of these 60 were 50 bytes or less. In another study [18] Kay et al. measured the TCP and UDP traffic on a FDDI LAN of Unix workstations in a university computer science department. They found that TCP message sizes are small: ....

Riccardo Gusella. A measurement study of diskless workstation traffic on Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, September 1990.

....applications (e.g. file transfer, remote job entry, database interrogation) an equally satisfactory modelling is not easily available. Furthermore, a number of studies have shown that in a LAN WAN environment the packet inter arrival times of the aggregate traffic clearly differ from exponential [10], 7] 11] On the other hand, a sound assessment of multiple access schemes for supporting the ongoing standardization activity on advanced mobile systems, 12] 13] needs to be based on realistic traffic characterization and scenarios. In order to evaluate the SIR capability to handle a ....

R. Gusella, "Measurement Study of Diskless Workstation Traffic on an Ethernet", IEEE Transactions on Communications, vol. COM-38, pp. 1557-1568, Sept. 1990

....of datagram traffic are bulk data transfer (FTP) and interactive applications (rlogin) Each datagram conversation is modeled by packet arrival rate and packet size. Several studies have been conducted to examine the datagram traffic characteristics on both local area and wide area networks [18, 19, 20, 21]. Instead of the popular Poisson model for packet arrival, Jain and others [20] have shown that the data transfer traffic follows a packet train model. The packet train model captures the burstiness of data transfer in which packets appear in groups or trains. The inter arrival time of the packets ....

Gusella R. A Measurement Study of Diskless Workstation Traffic on an Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, September 1990.

....the measurement system; no packet loss was detected. For more information on the measurement system, protocol analysis, and general traffic analysis, refer to [8] and to a previous study that examines measurements taken on another network, with characteristics similar to those of the Sun network [7]. Throughout the rest of this paper we will use client workstation to refer to a single user high performance personal computer as opposed to server workstation or file server , which is a machine that provides file space and file services. Some client workstations are diskless, while others ....

....measurements, and fitting. We mention here only some of works most relevant to the analyses presented in this paper. The interested reader will find additional references listed in these studies. Among related measurements studies, we would like to mention our previous study on Ethernet traffic [7] and a more recent work that provides some statistical characterization of the aggregate network traffic on an Ethernet [21] The properties of the hyperexponential distribution and the batch Poisson process that we have used are elementary. Most probability textbooks will suffice for background ....

Gusella, R., A Measurement Study of Diskless Workstation Traffic on an Ethernet, IEEE Transactions on Communications, (to appear) September 1990.

No context found.

R. Gusella. A measurement study of diskless workstation traffic on an Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, Sept. 1990.

No context found.

Riccardo Gusella. "A Measurement Study of Diskless Workstation Traffic on an Ethernet, "IEEE Transactions on Communications, 38:9, September 1990, pp. 1557 - 1568.

No context found.

GUSELLA, R. A Measurement Study of Diskless Workstation Traffic on an Ethernet. IEEE Transactions on Communications 38, 9 (Sept. 1990), 1557--1568.

No context found.

R. Gusella. A measurement study of diskless workstation traffic on an Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, Sept. 1990.

No context found.

R. Gusella. A measurement study of diskless workstation traffic on an Ethernet. IEEE Transactions on Communications, 38(9):1557--1568, Sept. 1990.

No context found.

Riccardo Gusella. A measurement study of diskless workstation traffic on an ethernet. IEEE Transactions on Communications, 38(9).

No context found.

GUSELLA, R. A Measurement Study of Diskless Workstation Traffic on an Ethernet. IEEE Transactions on Communications 38, 9 (Sept. 1990), 1557--1568.

No context found.

R. Gusella. A measurement study of diskless workstation traffic on an ethernet. IEEE Trans. on Communications, Vol. 38, No. 9, pages 1557--1568, Sept 1990.

No context found.

R. Gusella. A Measurement Study of Diskless Workstation Traffic on an Ethernet. In IEEE Transactions on Communications, volume 38, September 1990.

No context found.

R. Gusella. A Measurement Study of Diskless Workstation Traffic on an Ethernet. In IEEE Transactions on Communications, volume 38, September 1990.

No context found.

R. Gusella, A Measurement Study of Diskless Workstation Traffic on an Ethernet, IEEE Trans. Communications, vol. 38, (no. 9), Sep. 1990, pp. 1557-1568

No context found.

R. Gusella, "A Measurement Study of Diskless Workstation Traffic on an Ethernet", IEEE Transactions on Communications, Vol. 38, No. 9, pp. 1557-1568, September 1990.

First 50 documents  Next 50

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