Atom-Based Routing; Planning

Improving global internet routing by implementing atom-based routers.

8. Planning

First three months --- A Basic Atom-Based Router

In the first three months a basic atom-based router is implemented. This router is based on allocated atoms (Section 7.1). The work of the first period is carried out in the following steps:

1. Refining and Releasing Atom Computation Scripts To get Patrick Verkaik started, we begin with documenting Andre Broido's existing Perl scripts for atom computation. These scripts are subsequently released.
2. Implement Basic Atom-Based Router Next, we implement a basic atom-based router. The router is designed to work in the framework of Section 5.1, and can function as a transit or edge router. It uses existing router code (such as gated) wherever possible. Atoms are not yet computed; they are allocated (Section 7.1), since allocated atoms are more easily implemented.
3. Release We release the implementation.

Next three months --- Advanced Atom-Based Router

In the next three months an advanced atom-based router is implemented. In particular atom computation will be addressed. The following steps are taken:

1. Refine Atom Computation We review the above atom computation scripts with the aim of finding a scalable, realtime algorithm to compute atoms. One possible outcome is that atoms are initially computed as in the scripts (i.e. centralised and offline), but subsequently updated in a scalable, realtime way.
2. The improved algorithm is implemented, and incorporated into the atom-based router.
3. Evaluation We measure and evaluate the resulting atom-based router, comparing it to a standard BGP router. This could be done using an SSFnet simulation, with RouteViews as a data source to feed the simulation.
4. Release We release the implementation.

Notes:

• We intend to have experts looking over our shoulders from the start. Note that Tony Li (Procket Networks), Dave Ward (Cisco Systems), Curtis Villamizar (Avici Systems) and Dennis Ferguson (Juniper Networks) have already expressed interest. In this way we hope to have the results put to practical use for the benefit of the Internet community.
• Unfortunately, the precise start date remains unknown at this time, and depends on issues such as VISA application. We aim to get started at the beginning of September. For Patrick Verkaik to attend the workshop in Leiden in October, we would like Patrick to work from The Netherlands until the workshop, and to fly to San Diego after the workshop. The remainder of his six months will be carried out in San Diego.
• The planning above does not take into account vacation time.
• Four months into the project we review whether there should be an extension of another six months (Section 8.1).

8.1. Extension

Provided the outcome of the first six months is agreeable to all parties, we hope to extend the project for another six months. Ideas for the next six months are:

• Applying the atom concept to other domains. We expect atoms and our implementations to be applicable to other domains such as overlay networks (e.g. Peer-to-Peer and Ad-Hoc networks).
• Publishing the results.
• Creating specifications of the protocols that were derived or modified during the first six months. This involves removing any hacks (use of inappropriate or reserved IP header fields for example). The specification can be the basis of an RFC.