Skip to: Site menu | Main content

What do we plan to do?

Following are some ideas that we intend to pursue along with the endeavours being made at the moment.

You are most welcome to come up with your own ideas. If we find them to be in accordance with our objectives, we would be happy to have you as a part of team-maggie.

Measurement and Storage

The primary focus would be to evaluate new network monitoring tools and to extend the functionality of the PingER and IEPM-BW suites. Technologies will include relational databases, object oriented programming and service orientated architectures. Research and integration of other network monitoring systems will also be of importance. Following are the tasks planned under this:

Federation Techniques for Network Data Analysis
The sharing of network performance data is vital in understanding the network as a whole. Projects such as perfSONAR and Internet2 piPES
provide in-depth router performance statistics, the understanding and correlation of which is of great value. Similarly Monalisa provides advanced aggregation and presentation techniques. Here the work will focus on developing such technologies in accordance to the work of IEPM to aid the effective sharing and analysis of the data.

Evolution of IEPM-BW and PingER Architectures
An important part in being able to provide network performance data lies in the ability for the IEPM-BW and PingER architectures to efficiently and accurately produce network performance data. However, as with all designs, the architectures need to be expanded upon to provide additional features, core systems redesign for efficiency and a modular component layout. Here the aim would be to evolve the architectures of IEPM-BW and PingER as per the change in requirements.

Monitoring of High Performance Networks
As PERN2 will be 10Gbps, there will be new challenges in understanding the effects of increased capacity on existing measurement tools and techniques. This task will evaluate and report on such issues and aid the selection of new measurement tools and techniques for future monitoring needs.

Data Analysis and Presentation

Using the vast amounts of network monitoring data already available, here the roles would be the mining and analysis of network performance trends and developing innovative and useful forms of data visualization. Examples include the developing web sites to accommodate and present such information with special attention to the unification and inter-machine communication of network data using XML. Technologies will include Perl/Python CGI, Application Servers and Service Orientated Architectures.

Development of a Web Service Front-End for Pinger and IEPM-BW
The intercommunication of network data is essential to provide access to network performance information. An essential part of this project will be to develop and implement standard XML schemas from which network performance consumers can query our extensive databases.

Visualization and Analysis of Internet Topologies
Due to the transparent nature of Internet design, it is often difficult to make sense of network path data. This project will focus on creating visual aids from which network performance data can be 'browsed' in a logical and consistent manner.

Netflow Data Analysis and Presentation
Passive monitoring will become important in the future as network speeds increase - as the action of active network testing more and more difficult. Netflow is a standardized information format from which real network traffic information from routers can be processed and analyzed. This project will investigate into the creating and presenting Netflow data with emphasis on comparison between active and passive measurement techniques.

Problem Isolation and Alerting

Using advanced mathematical techniques and algorithms, efforts would be made to isolate performance problems at both the network and compute levels. Techniques such as Holt-Winters will be used to automatically detect issues in the network and alert relevant parties of such trouble.

Anomaly Detection Methods & Analysis of Alerting Techniques
Some basic functionality is already in place to provide alerts within the IEPM-BW framework when performance changes occur. This project will both provide a more generic method to the automated analysis of network performance data, and also investigate and implement advanced algorithms for such analysis and reporting.

Event Isolation and Diagnosis Techniques
Currently, diagnosis of network problems requires much administrative manpower and time to determine the cause of the problem(s). The need to quickly and automatically identify and diagnosis network problems to dramatically decrease the period that users experience degraded performance. This project aims to automatically gather relevant information, and use heuristical analysis to enable location (e.g. network or host issues), identification (narrowing down of the cause), and verification (double-checking against maintenance tickets etc.) of events.

Analysis of Forecasting Techniques
The prediction of network performance will greatly aid large scale data replication applications. However, such techniques are currently not very well understood. This project will investigate and report upon techniques to provide reliable forecasts of network performance based upon past performance.