Manet meeting minutes from the 45th IETF Wednesday, July 14, 1999 1300-1500 The mobile ad hoc networking (manet) WG held one meeting in Oslo on Wednesday, July 14 at 1300-1500. The session was chaired by Joe Macker (the co-chair Scott Corson could not attend). The meeting began with a short introduction and agenda review by Joe Macker. The agenda was roughly organized into two parts. First, recently or "soon to be" updated IDs were scheduled for review by the corresponding authors. Recent ID updates of Dynamic Source Routing (DSR) and Ad Hoc On-Demand Distance Vector (AODV) drafts were reviewed. In addition, protocol improvements and performance analyses for the Cluster-based Routing Protocol (CBRP) and Optimized Link State Routing (OLSR) were described. In the second half of the meeting, additional manet architecture and issues presentations were scheduled. These presentations included topics such as; mobile IP and manet integration, manet protocol energy consumption, and hierarchical manet architectures. Existing Manet Protocol Review and Status The first presentation was from the CMU Monarch team: Dave Johnson, Josh Broch, and Dave Maltz. An updated version of the DSR ID has been recently provided within manet. Josh Broch began by presenting a review of the present DSR manet addressing approach. In the DSR addressing approach, a single IP address is used per node with indexes to identify multiple interfaces. At present for external communications, a single address space is used for all nodes in a manet and the potential for increasing scalability through the use of multiple manet substructures was briefly discussed. The chair provided a related addressing comment recommending consideration for future support of subnet masking in DSR route collection. CMU appeared to agree that this extension was desirable and indicated it would be accommodated in a future revision. Dave Maltz next discussed continuing work by CMU in providing openly available manet-related tools within the Berkeley ns2 simulation environment. This talk focused on a manet emulation capability that allows interaction of manet simulation models with live, implemented systems. This will allow for easier and more comprehensive testing and development in the future. The emulation work leverages off initial ns2 emulation work done by Kevin Fall under the VINT project. Encouraging validation results were presented comparing the simulation and emulation modes. The ability to introduce and evaluate real hardware and applications within the emulation environment was discussed as an additional benefit. New CMU software ports to ns2 were anticipated to be available in about 2 months. Jiang Mingliang followed the CMU presentations with an update of recent CBRP modifications and evaluation results. First, a number of protocol modifications were discussed. A local repair mechanism has been added to CBRP to improve the packet delivery ratios. A simulation of CBRP has been developed based upon the CMU ns2 manet extensions. Results showed that the CBRP improved the packet delivery ratio relative to DSR for large networks (e.g., 150-200 nodes). It was claimed that CBRP would be a good choice for large scaled manet scenarios. One group comment requested that additional overhead results be provided in packet count so that these could be directly compared with previous DSR results. It was agreed that results could be provided in both forms equally well and that this would be done in the future. Charlie Perkins provided an update and overview of ongoing AODV work. An updated ID has been provided which covers improvements to the protocol with contributions from Elizabeth Royer and Samir Das. Extensions discussed included the following: service location functionality, expanding ring search, and multicast algorithm changes. Service location extensions to AODV were discussed and were based upon RREP/RREQ by protocol and port number. A long-lived association between the service and the IP address was assumed here. In addition, as an alternative to the "1-hop then infinity" searching approach a higher fidelity expanding search capability has been added. Questions remain regarding the best expanding ring search algorithm to use and more work is needed here. Also, a multicast modification has been added to AODV regarding the process of merging disjoint trees. Other changes are in the ID as well, including a broadcast algorithm improvement and additional flags to improve the formation of trees. There was a group comment questioning the rationale of doing service location in AODV vice using SLP. The group seemed to agree that there was general interest in mobile location protocols applied to manet and that other areas may also be of interest including anycast,etc. The SLP issues require more WG discussion. Amir Qayyam provided an update to the OLSR protocol and discussed some initial evaluation results. A power saving modification was discussed which allows OLSR nodes to go into sleep mode. A modification to reduce the amount of overhead required to disseminate multipoint relay set information was also described. Initial simulation efforts of OLSR and a skeleton DSR were described. This simulation used an internal INRIA simulation environment to take advantage of HIPERLAN MAC layer models. A modified IMEP was developed to combine BEACON and HELLO into a single HELLO message. This was claimed to have reduced OLSR overhead within the simulations. There was no mobility in the initial simulation runs presented. A group comment questioned the present value of the results due to this lack of mobility. It was explained that these were preliminary results and that more detailed simulations would follow. Additional Manet Topics and Proposals Laura Sweeney provided a presentation of energy consumption and manet protocols. The energy cost at both the sender and the receiver was discussed. Laura provided a quick overview of the energy consumption models developed and also discussed the relationship to empirical studies done on Wavelan hardware by others. Laura described that minor modifications were made to the ns2 extensions and simulation results were collected. Due to a lack of time, not all results could be presented in detail and it was recommended that results be provided electronically for further WG perusal and discussion. Jari Malinen provided a presentation of manet studies and work that are ongoing at the University of Helsinki. They are presently working on a combination of manet and mobile IP approaches for their envisioned mobile architecture. More results from this work will be presented to the working group in the future. Hong Jiang presented a brief discussion on a draft framework for what was termed " An Architecture for High Speed Manet". The definition included a mobile architecture consisting of "high-speed" links (OC3 point-to-point technology was mentioned), wide area coverage (100s of miles), and a large number of nodes (100s of backbone nodes and potentially 1000s of subscribers). It was claimed that mobility management and routing should be separate. There was an indicated requirement to provide QoS services and provide distributed operation and configuration. This architecture appeared to be a description of a pseudo-cellular network in which a wireless backbone of access point nodes formed a manet and were potentially mobile. End nodes were not envisioned to be routers and mobile users gained access to the mobile infrastructure through these access points (which may be mobile). There was interest in applying present or new manet technology to this problem. The group raised the following question. Where does the mobile routing function reside and is it envisioned to use manet technology? The answer was that the backbone nodes are envisioned to provide a self-organizing, mobile routing infrastructure and may take advantage of existing manet routing protocols or something new. The group took issue with the term "high speed" and recommended some other adjective be used to describe this application area. It was recommended that the authors consider different terminology and provide more details regarding a specific protocol approach prior to posting an Internet Draft.