Overview

In the last decades Internet has grown from an academic experiment with several small attached networks to a highly interconnected heterogeneous system that spans several continents. Recently it is a network of networks that consists of millions of private and public, academic, business, and government networks of local to global scope. Besides the Internet's expansion the growing number of users and applications generate huge and more complex network traffic to be handled which possess many challenges for network operators and the network itself. As a consequence traffic control, forecasting, performance analysis and monitoring are becoming fundamental issues for network operators and interesting targets for researchers as well.

To determine the key performance metrics needed to analyze network behavior and network traffic, numerous independent network measurement infrastructures and testbeds have been developed and deployed all over the world. These infrastructures aim at helping researchers to examine many interesting aspects of the Internet like network topology, traffic behavior, one-way and queuing delay fluctuations or routing policies. Nevertheless, the way to use them is very different and sometimes not too comfortable. In general, it splits into the following key steps: first we write a measurement script then we upload it to the measurement nodes (via Web or a direct terminal connection). After that we execute it and finally we collect the results and optionally store them in a database.

In addition, we have to mention that some ISPs provide publicly accessible measurement services, which are very popular among Internet users thanks to their Web based accessibility. Their network measurement capabilities are, however, very limited and not sufficient for the research community. One of the open question in network research is how the flexibility of the major network measurement infrastructures can be combined with the general accessibility and popularity of these lightweight Looking Glass services.

We have developed a Web Service based approach, called SONoMA, for building an integrated architecture for network measurements that is scalable, adaptable and open for scientists and other network developers, while its functionalities can be easily accessed through a standardized interface. The Service Oriented Architecture (SOA) is a very popular principle in system design and integration concerning business applications. Naturally the key components of this principle can be used in the design of network measurement architectures as well.

SONoMA is a common and extensible network measurement framework to define distributed network experiments. To perform a measurement using this system there is only one thing to do: based on a predefined interface description prescribe what you want to measure and then the framework will ensure to deliver the required measurement data. Nevertheless, the above request will be disassembled as individual executable tasks in the background. Each task will be performed on a proper set of the measurement nodes in a completely distributed manner. Whilst the results are forwarded back to the user they are in parallel automatically stored in our public database, which is called the Network Measurement Virtual Observatory (VO).

© 2016 SONoMA: A Service Oriented Network Measurement Architecture