New source-level models for aggregated HTTP traffic and a design for their integration with the TCP transport layer are built and validated using two large-scale collections of TCP/IP packet header traces. An implementation of the models and the design in the ns network simulator can be used to generate web traffic in network simulations.

Real production applications ranging from enterprise applications to large e-commerce sites share a crucial but seldom-noted characteristic: The relative frequencies of transaction types in their workloads are nonstationary, i.e., the transaction mix changes over time. Accurately predicting application-level performance in businesscritical production applications is an increasingly important problem. However, transaction mix nonstationarity casts doubt on the practical usefulness of prediction methods that ignore this phenomenon. This paper demonstrates that transaction mix nonstationarity enables a new approach to predicting application-level performance as a function of transaction mix. We exploit nonstationarity to circumvent the need for invasive instrumentation and controlled benchmarking during model calibration; our approach relies solely on lightweight passive measurements that are routinely collected in today’s production environments. We evaluate predictive accuracy on two real business-critical production applications. The accuracy of our response time predictions ranges from 10 % to 16 % on these applications, and our models generalize well to workloads very different from those used for calibration. We apply our technique to the challenging problem of predicting the impact of application consolidation on transaction response times. We calibrate models of two testbed applications running on dedicated machines, then use the models to predict their performance when they run together on a shared machine and serve very different workloads. Our predictions are accurate to within 4 % to 14%. Existing approaches to consolidation decision support predict post-consolidation resource utilizations. Our method allows application-level performance to guide consolidation decisions. 1.

This tutorial article discusses the role of network traffic measurement in the design, testing, and evaluation of Internet protocols and applications. The article begins with some background information on Internet traffic measurement, and then proceeds to discuss the “tools of the trade”, including examples of both hardwarebased and software-based approaches to network traffic measurement. The article concludes with a summary of the main observations from the past fifteen years of network measurement research, along with pointers to the relevant literature for more information. Keywords: Network traffic measurement, workload characterization, Internet protocols, TCP/IP, performance 1

Abstract not found

The burstiness of Internet traffic was established in pioneering work in the early 1990s, which demonstrated that packet arrival times are not Poisson, and packet and byte counts in fixed-length intervals are long-range dependent [17, 20]. Here we demonstrate that these results are one end of a continuum of traffic characteristics. At the other end are Poisson behavior and independence. Our study focuses on packets, what devices actually see; we study the statistical properties of packet inter-arrival times and packet sizes. As the traffic load increases --- that is, as the number of simultaneous transport connections increases --- arrivals tend to Poisson and sizes tend to independence. More specifically, long-range dependence of inter-arrivals and sizes decreases to independence, and the marginal distribution of inter-arrivals tends toward exponential; this happens (1) through time on a single link as the load increases due to daily variation, or (2) at a single point in time as the load increases going from lightly loaded links at the edges of the Internet to heavily loaded links at the core. Convergence is rapid; the packet traffic gets quite close to Poisson and independent loads far less than the maximum we observe.

Our goal is to design a traffic model for non congested Internet backbone links, which is simple enough to be used in network operation, while being as general as possible. The proposed solution is to model the traffic at the flow level by a Poisson shot-noise process. In our model, a flow is a generic notion that must be able to capture the characteristics of any kind of data stream. We analyze the accuracy of the model with real traffic traces collected on the Sprint IP (Internet Protocol) backbone network. Despite its simplicity, our model provides a good approximation of the real traffic observed in the backbone and of its variation. Finally, we discuss the application of our model to network design and dimensioning.

The utility of simulations and analysis heavily relies on good models of network traffic. While network traffic constantly changing over time, existing approaches typically take years from collecting trace, analyzing the data to finally generating and implementing models. In this paper, we describe approaches and tools that support rapid parameterization of traffic models from live network measurements. Rather than treating measured traffic as a timeseries of statistics, we utilize the traces to estimate end-user behavior and network conditions to generate applicationlevel simulation models. We also show multi-scaling analytic techniques are helpful for debugging and validating the model. To demonstrate our approaches, we develop structural source-level models for web and FTP traffic and evaluate their accuracy by comparing the outputs of simulation against the original trace. We also compare our work with existing traffic generation tool and show our approach is more flexible in capturing the heterogeneity of traffic. Finally, we automate and integrate the process from trace analysis to model validation for easy model parameterization from new data.

Optical Burst Switching (OBS) is considered as a promising switching technique for building the next generation optical Internet. In OBS networks, one important issue is how the performance will be affected by bursts assembled from packets, which is the basic transmission unit in OBS. In this paper, we study the fundamental statistic properties such as the burst length distribution, inter-arrival time distribution, as well as correlation structure of assembled burst traffic from burst assembly algorithms. From both theoretical and empirical results, it is demonstrated that after the assembly, the traffic will in general approach the Gaussian distribution. In particular, the variance of assembled traffic decreases with the increase in the assembly window size and the traffic load. However, the long range dependence in the input traffic will not change after assembly.

In this paper, we propose SFD algorithm to reduce the user-perceived web response time (i.e. web latency) . This algorithm gives short flows preferential treatment and penalizes long flows. We implement SFD algorithm as a simple differentiated services policy and evaluate its performance in simulation. We find that the transmission latency of short flows and the response time to retrieve representative web pages are both reduced by about 30%. Using web traces we demonstrate that 99% web pages would be transferred faster. SFD also bounds the penalty to long flows. We further evaluate how different schemes trade-off the performance between short and long flows.

As the active connection load (ACL) on an Internet link increases, the statistical properties of packet inter-arrivals and sizes change due to increased statistical multiplexing of packets from different connections. Chief among these results is that the long-range dependence of the inter-arrivals and sizes goes locally to independence. The results are based on (1) the mathematical theory of superposition of marked point processes, and (2) empirical study of 3026 packet traces, each 5 minutes or 90 seconds in duration, from 6 monitors on Internet links ranging from 100 mbps to 622 mbps. An understanding of packet inter-arrivals and sizes is important because it is packets that devices must send and receive, and the burstiness of traffic as seen by the devices is determined by the statistical characteristics of these two variables. The results for inter-arrivals and sizes do not conflict with previous well known results about the statistical properties of packet and byte counts in fixed time intervals, but the counts do have a different change in statistical properties with the ACL. 1.