E2E piPEs

End-to-End Performance Initatives Performance Environment System

MOST RECENT PRESENTATION:	piPEs Update Joint Techs SLC summary [PPT] (February 2005) piPEfiitter BoF Joint Techs SLC summary [PPT]
PROJECTS USING PIPES DATA:	http://e2epi.internet2.edu/e2epipes/user-dir.html

NOTE: The piPEs work was merged into an International effort, called perfSONAR, in mid-2005. Continuing development of the framework has occurred since that time. For more information, please see: www.perfsonar.net

The E2E piPEs framework is the basis for much of the work of the Internet2 End-to-End Performance Initiative in the coming year. In its final form, piPEs will be able to indicate performance capabilities and locate performance problems along the path between two computers connected by the Internet2 Abilene network, participating campuses, regional networks, and gigaPoPs. In turn, this will significantly improve the likelihood that advanced Internet applications can operate at peak performance and thereby advance the productivity of academic researchers. At this time, E2E piPEs has four goals:

Enable end-users and network operators to determine E2E performance capabilities, locate E2E problems, and contact the right person to get an E2E problem resolved.
Enable remote initiation of partial path performance tests.
Make partial path performance data publicly available.
Be interoperable with other performance measurement frameworks.

The goal of the architecture of the piPEs framework (shown in the figure below) is to determine the performance characteristics of the complete path by aggregating information about the segments that make up the path; problematic partial paths can be identified and reported, with supporting data, to the appropriate network administrator.

A battery of regularly scheduled active tests provides information on loss, jitter, throughput, and one-way latency data. If the necessary data is not included in the database, a test can be scheduled on demand. For example:

The end user is experiencing problems; he requests information, specifying end points and type of application.

The request is scrutinized to determine whether the requestor has the authority to make such requests and whether the two end-points share the same tool. If the answer to either of these is “no”, the request is rejected.
If the answers are “yes”, piPEs checks the database to determine if the request can be met by regularly scheduled tests. If not, a test is scheduled on-demand.
The test results are gathered and stored in a distributed centralized database. Then, piPEs analyzes the data and provides the user: a) the likely place to start looking for failure, b) contact information for the individual responsible for that administrative domain, and c) data to provide to the contact with sufficient cause to investigate the complaint.

The aim of this system is to reduce the “signal to noise ratio.” NOTE: piPEs, as described above, is the final product; at this time, work on the project is focusing on the database, web-based display engine, the analysis engine, and performance measurement points. The regularly scheduled tests include latency (OWAMP), bandwidth (BWCTL), and routing information (traceroute).

The initial deployment, which includes the Abilene backbone network and a few campuses, has been demonstrated at several workshops, including: TIP 2004, GNEW 2004, the Transatlantic Performance Monitoring Workshop, and the CANARIE-GEANT-Internet2 Lightpath Workshop. These demonstrations have included transcontinental, transpacific, and transatlantic paths; they have also demonstrated the ability of other tools to consume the data provided by piPEs – both NLANR’s The Performance Advisor and the HENP-funded MonALISA project are able to display data collected by piPEs.

Preliminary steps to deploy piPEs at the NC-ITEC (NC-State) and ITEC-Ohio (OSU) have begun. The CENIC network is deploying a similar backbone network infrastructure to the AMI (step 1 in the piPEs framework); when the infrastructure is complete, they will be ready to implement piPEs both in their own network and linked to the Abilene network. Current deployment efforts are shown below:

A group of collaborators, known as piPEfitters, are working on various aspects of the project. Send your comments or questions to eboyd@internet2.edu. or subscribe to our mailing list.

Note: The development of piPEs has benefited from extensive conversations with colleagues from the Internet2 E2Epi TAG team, the DAST/NLANR team, SLAC researchers, the NLM Visible Human Project, HENP and VLBI researchers, the Merit team, the Internet2 ITECs, DANTE, the University College of London, and many others. Their invaluable comments and insights are much appreciated.