NETWORK PREFORMANCE - SELF-SIMILARITY NATURE OF ETHERNET TRAFFIC

INTRODUCTION

This self-similar or fractal -like behavior of aggregate Ethernet LAN traffic is very different both from conventional telephone traffic and from currently considered formal models for packet traffic (e.g., pure Poisson or Poisson-related models such as Poisson-batch or Markov-Modulated Poisson processes). The main objective of this project is to establish in a statistically rigorous manner the self-similarity characteristic of the very high quality, high time-resolution Ethernet LAN traffic measurements.


NATURE OF ETHERNET TRAFFIC


Packet streams generated from individual sources superimpose multimedia traffic in Broadband Integrated Service Digital Network (BISDN). Each source is expected to alternate between ON and OFF periods and generate packets in ON period. By stochastic modeling, such an ON/OFF nature is best captured by a binary process. The characteristics of a generic binary process are described by three random variables:

The simplest binary process is 2-state Markov chain, which is regarded as the first order approximation model since it only match the average behavior of the three random variables. That is, both ON and OFF periods are assumed to be geometrically distributed and the access rate in on period is constant. Below are 2_state MC to characterize the first-order properties of each binary source, described by

The average access rate is defined by the average input rate of each source in steady state. The peak access rate is the average source access rate in ON period. The ratio of the peak access rate to the average rate characterized the burstiness of each individual source. The rate time-varying scale can be defined as the cumulative correlation of source access rate at different time intervals. It is anticipated that the multimedia traffic will be characterized by great diversities of traffic behavior. Indeed, each of these properties of individual sources ranges from a few KBPS of interactive data to a few hundred MBPS of high definition TV. The peak access rate of a packetized video source, as shown by some statistic, can be one order of magnitude higher than its average value access rate. Further, the time span between the fastest and the slowest variation of multimedia traffic is at least by several orders of magnitude, ranging from an interactive random data generator to a deterministic traffic stream.

LAPSED TIME CAUSED BY ETHERNET TRAFFIC

With the growing Web and other Internet traffics plus multimedia services such as video-on-demand on the horizon, revamping local-network access has become a top priority. The problem posed by Internet traffic is that its traffic dynamics (burstiness and long holding times) are radically different from voice telephone services (many calls of comparatively short duration). Local exchange carriers (LECs)

PROPERTIES OF SELF-SIMILARITY

Self-similarity means that the traffic has similar statistics properties at a range of time scales: milliseconds, seconds, minutes, hours, even days and weeks. This has several important consequences. One is that you cannot expect that the traffic will "smooth out" over an extended period of time; instead, not only does the data cluster, but the clusters cluster. Another consequence is that the merging of traffic streams, such as is done by a statistical multiplexer or an asynchronous-transfer mode (ATM) switch, does not result in a smoothing of traffic. Again, multiplexing bursty data streams tends to produce a bursty aggregate stream.

WHERE IS SELF-SIMILARITIES FOUND?

Self-similarity applies not only to Ethernet traffic or indeed to local-area network traffic but also in ATM traffic, compressed digital video streams, Signaling System Seven (SS7) control traffic on networks based on the integrated services digital network (ISDN), Web traffic between browsers and servers, and much more.

AWEARNESS DUE TO SELF-SIMILARITY

The implications of self-similarity of data traffic are startling and revealed its importance. For example, the whole area of buffer design and management requires rethinking. In traditional network engineering, it is assumed that linear increases in buffer size will produce nearly exponential decreases in pack loss and that an increase in buffer size will result in a proportional increase in the effective use of transmission capacity. With self-similarity traffic, these assumptions are false. The decrease in loss with buffer size is far less than expected, and a modest increase in utilization requires a significant increase in buffer size.

OTHER AFFECT ON NETWORK DESIGN

Other aspects of network design are also affected. With self-similar traffic, a slight increase in the number of active connections through a switch can result in a large increase in packet loss. In general, the parameters of a network design are more sensitive to the actual traffic pattern than expected. To cope with this sensitivity, designs need to be more conservative. Priority scheduling schemes need to be reexamined. For example, if a switch manages multiple priority classes yet does not enforce a bandwidth limitation on the class with the highest priority, then a prolonged burst of traffic from the highest traffic could keep the other classes from using the network for an extended period of time.

INGREDIENCE OF THE PROJECT

In order to venture into the field of network performance. It is important to establish a brief understanding of an average network system. When computers are distributed over a localized area-such as a building or campus-the network used is known as a local area network (LAN), when the computers are distributed over a wider geographical area-such as a country - the network is known as a wide area network (WAN), when the complete network comprises an interconnected set of local and wide area networks, it is known as an internetwork. Figure below shows us how an office LANs setup is established.

Refer to figure in page 4 of SQ757 LAN lecture notes


HOME