Tuesday 25 March 17:00 - 18:30 Colorado Hall E
BOF 1: IRTF Network Management Research Group
Chair: Prof. Jürgen Schönwälder, Universität Osnabrück, Germany

The Network Management Research Group (NMRG) of the Internet Research Task Force (IRTF) was formed in 1999 to provide a forum for researchers and technology experts to explore new technologies for the management of the Internet. So far, the NMRG has held 12 face-to-face meetings, discussing new technologies, approaches and ideas in technical details.

The NMRG BOF allows interested people to get in touch with the NMRG and to discuss previous, current and future NMRG work items or meeting topics in an informal setting.

Tuesday 25 March 17:00 - 18:30 Colorado Hall F
BOF 2: GRID Management: Network and Data Link (Layer 3 and 2: L32) Layer Perspective

Chair: Dr. Masum Z. Hasan, Cisco Systems

GRID is an active research and development area with focus on distributed, parallel, and networked computing. What used to be a purely academic research area, especially in the parallel system community has evolved into something encompassing anything that is distributed, parallel and networked. The concept also got its way into commercial arena with all the associated hype.

The overwhelming majority of industrial, commercial and academic activities are geared towards what in IOS-OSI layering terms belong to layer 7 (L7). That is, majority of work is focused on applications, servers, and middleware. There has been some level of work in terms of L32, but much more is needed. This is especially important in building GRIDs in a (what I'd call) heavily networked environment.

A L7 GRID is generally built as an overlay network on top of L32 networks. The functioning, configuration and behavior of the underlying network is assumed to be independent of a GRID network. This concept of transparency is good. But for building better and new GRIDs and GRID applications it is important that some level of interactions with the L32 be provided. What is needed is an L32 framework or interface to which the GRID middleware can plug into. The framework can be based on typical Network Management FCAPS.

The framework can be used in a number of ways. Let me provide one example here (more examples and further details will be provided in BOF prezo). Let's consider building overlay GRID networks with appropriate QoS between GRID points. For example, one may build a Game GRID with communicating game servers distributed worldwide. Two clients across the globe playing with each other may connect to their respective local (nearest) servers. The servers communicate with each other while the clients play, and certain QoS may need to be provided on the communication channel between servers. Client to server QoS (bandwidth, delay, jitter) may also be needed. The GRID overlay network can be built by providing appropriate network connections with specified QoS between the servers. Once the server points (on the GRID) are decided, the GRID overlay network configurator or middleware may interact with L32 interfaces instructing it to provide appropriate QoS (bandwidth, delay, jitter) between GRID points (servers). The L32 interfaces translate the requests to L32 QoS and configure appropriate routers or switches. For example, depending on underlying network MPLS TE tunnels or ATM PVCs with appropriate QoS may be configured between GRID points. (Further detail will be provided in BOF prezo).

Note however though, that because of wide varieties of L32 technologies, and vendors, it will not be easy to define such a framework. Some level of abstractions may need to be provided.

Further examples and concepts (such as GRID aware routing, L32 aware scheduling or load-balancing, GRID VPN, etc.) will be discussed in the BOF.

This BOF will attempt to bring interested parties together to work jointly in this particular area. To define such a framework interactions between L7 GRID and L32 experts are required.

Wednesday 26 March 17:00 - 18:30 Colorado Hall D
BOF 3: Informal intro and questions about cfengine

Chair: Prof. Mark Burgess, Oslo University College, Norway

Cfengine is a (GNU) project from Oslo University College for autonomously managing Unix and Windows hosts without the use of SNMP. It uses a medium-level policy based language that has a property of "convergence to stability". Cfengine is installed on hundreds of thousands of systems around the world.

Wednesday 26 March 17:00 - 18:30 Colorado Hall E
BOF 4: Dynamic Policy-based Control in Distributed Networks

Chair: Dr. Petre DINI, Cisco Systems, Inc.

BOF Intended audience: Specialists in Systems Control and Management

The timing has never been better for service providers to expand their portfolio of offerings, from high speed backbone, to IP VPN's and managed access. Enterprises and small businesses are increasingly evaluating their outsourcing options, and looking for a range of converged voice, video and data services. As the converge occurs, the challenges of operating and managing an integrated network intensifies. A resilient, flexible operational infrastructure becomes the critical success factor in offering a granular and subscriber self-managed portfolio.

The BOF will focus on several of the operational challenges of a "ubiquitous network of networks", and offer a strategy, vision, and concrete policy-enabled solutions toward adaptive and self healing networks of the future. Emphasis will be given to the increasing function of the programmable layers, a policy-enabled management and service paradigm for enabling network and service convergence, feature interaction free, and robust through failure-avoidance.

The intention is to identify and form a specialized work group of specialists interested to cooperate on the topics related to the exposed BOF Intent.

Wednesday 26 March 17:00 - 18:30 Colorado Hall F
BOF 5: SAGE: The Organization for Professional System Administrators

Chair: Rob Kolstad, Executive Director, SAGE

SAGE, a special technical group of USENIX, is expanding its activities directed towards enhancing the professional stature of the field of "System Administration". SAGE's Executive Director, Rob Kolstad, will give a 30 minute presentation on some of SAGE's current activities and goals (including the 'Book of Knowledge', 'Code of Ethics', and college curriculum committee) and then open the floor for questions and discussion. Contact kolstad@sage.org for details.