Packet Inter-arrival Time Distributions Research Articles

Taming the Elephants: Affordable Flow Length Prediction in the Data Plane

Machine Learning (ML) shows promising potential for enhancing networking tasks by providing early traffic predictions. However, implementing an ML-enabled system is a challenging task due to network devices limited resources. While previous works have shown the feasibility of running simple ML models in the data plane, integrating them into a practical end-to-end system is not an easy task. It requires addressing issues related to resource management and model maintenance to ensure that the performance improvement justifies the system overhead. In this work, we propose DUMBO, a versatile end-to-end system to generate and exploit early flow size predictions at line rate. Our system seamlessly integrates and maintains a simple ML model that offers early coarse-grain flow size prediction in the data plane. We evaluate the proposed system on flow scheduling, per-flow packet inter-arrival time distribution, and flow size estimation using real traffic traces, and perform experiments using an FPGA prototype running on an AMD(R)-Xilinx(R) Alveo U280 SmartNIC. Our results show that DUMBO outperforms traditional state-of-the-art approaches by equipping network devices data planes with a lightweight ML model. Code is available at https://github.com/cpt-harlock/DUMBO.

Application-Oriented Traffic Modeling of WiFi-Based Internet of Things Gateways

Many Internet of Things (IoT) devices generate relatively small-sized data and have limited energy supply. These two factors limit their ability to connect directly to cloud servers through a wireless backbone network without imposing a burden on this network in providing efficient data transfer. In this article, we consider an IoT network architecture where a number of different IoT devices send their data wirelessly to an IoT gateway (or a fog node) via a WiFi network. We focus on characterizing incoming traffic patterns to the gateway for three typical IoT applications with real-time and nonreal-time data transfer requirements, such as video surveillance, smart city, and e-healthcare. Our study is based on generating real IoT traffic traces in a lab environment from various sensors and devices for the aforementioned applications and employing these traces to emulate a network of IoT nodes connected to a gateway via WiFi. In the conducted experiments, different homogenous and nonhomogeneous traffic patterns of the selected applications are examined for synchronized and unsynchronized data sources. Based on our empirical data, the experimental results reveal that the packet interarrival time distribution at the gateway is close to generalized Pareto distribution for homogeneous eHealth and smart city traffic, whereas the Weibull distribution is the nearest to model the empirical packet interarrival time for the rest of the examined traffic patterns. Moreover, we show that employing the experimental findings to analyze the delay performance of connecting the gateway to the cloud, given certain backbone network resources, leads to accurate results.

Packet Inter-arrival Time Distribution in Academic Computer Network

The article presents the statistical analysis results of network packet inter-arrival time distribution in academic computer network. Most popular transport protocols TCP and UDP are addressed in the research. Data was gathered using NetFlow protocol. Network traffic was divided into sections according its direction and usage trends, then packet inter-arrival time distributions were found. Kolmogorov-Smirnov test was used to evaluate goodness-of-fit of packet inter-arrival time distributions and it was determined, that Pareto Second Kind distribution fits the majority of the experimental distributions. DOI: http://dx.doi.org/10.5755/j01.eee.20.3.6683

Detecting suspended video streams through variance-time analysis

To detect suspended video streams, an algorithm is proposed which analyzes traffic data to sense changes in variances. The objective of the algorithm is to help restart suspended streams as quickly as possible in cooperation with network administrators, mechanisms to change routes, etc. The algorithm handles traffic data of a number of streams all together to detect one or more suspended streams without using the information on the number of streams and on each stream's data rate. It works with the same parameter values for both LAN and WAN traffic streams, even if their packet interarrival-time distributions are different. Experiments with 41 live TV streams showed that suspension is detectable when the traffic rate decrease caused by the suspension is 1.6%.

Throughput of Unslotted ALOHA Channels with Arbitray Packet Interarrival Time Distributions

An unslotted ALOHA channel with dissimilar users and each user having an arbitrary distribution of packet interarrival time, is analyzed for its throughput. It is shown that the throughput behavior of an unslotted channel with a very large number of users is the same as that of Abramson's Poisson channel. When each user transmits many packets almost in succession and then remains silent for a long period, the throughput of the unslotted channel is shown to approach that of the slotted channel. For unslotted channels in which the dead time between successive packet transmissions due to each user exceeds twice the packet duration, the throughput is shown to be independent of the nature of the packet interarrival time distribution of any user, depending only on the average rate at which users transmit packets.