Mobile Metropolitan Ad Hoc Networks Mobileman Mobileman Technical Evaluation Project Funded by the European Community under the " Information Society Technologies " Programme

The aim of this deliverable is to present the validation results of the architecture, protocols and services designed for the MobileMAN project. This deliverable therefore represents the complement of Deliverable D13. Specifically, D13 presents the MobileMAN architecture and protocols, while in this deliverable the solutions we devised are validated. Whenever possible, our evaluations are based on measurements from real smallmedium-size testbeds. In addition, simulation results are used to study large scale networks and/or complex mobility scenarios. Results are presented by following a bottom up approach from wireless technologies up to the applications. The deliverable also investigates the behavior and performance of a medium-scale MobileMAN network made of (up to) 23 nodes. Project funded by the European Community under the “Information Society Technologies” Programme (1998-2002) MOBILEMAN IST-2001-38113 October 2005 Deliverable D16 3 of 222 Summary The aim of this deliverable is to present the validation results of the architecture, protocols and services designed for the MobileMAN project. This deliverable therefore represents the complement of Deliverable D13. Specifically, D13 presents the MobileMAN architecture and protocols, while in this deliverable the solutions we devised are validated. There are two main approaches in system evaluation: measurements on real testbeds and analytical/simulation modeling. The performance study of mobile ad hoc networks is a complex task that cannot be addressed by using a unique performance technique but requires a careful mixing of measurements on real testbeds with simulation/analytical studies. Whenever possible we constructed smallmedium-scale testbeds to validate our solutions by taking into consideration real scenarios. Modeling studies have been extensively used in the protocols’ design phase, and to study the MobileMAN system behavior in complex scenarios that are very difficult (if not impossible) to be studied by prototypes. In the latter case, to develop and solve our simulation models we used simulation tools. In this deliverable we present a comprehensive overview of all the activities we performed to validate the solutions we developed for the MobileMAN environment. The deliverable follows a bottom up approach from wireless technologies up to the applications. The deliverable ends with a section reporting the experimental evaluation of a MobileMAN medium-size mobile-ad-hoc network (up to 23 nodes) which integrates the solutions we have developed. Specifically, in Section 1, after discussing the performance modeling techniques, we present the characteristics of the simulation framework, which extends the Network Simulator NS-2 (v. 2.27) with a cross-layer interface (XL-interface) that standardizes vertical interactions among protocols according to the MobileMAN cross-layer architecture. This simulation framework has been used in successive sections to validate our cross-layer solutions (e.g., see Reliable Forwarding and crosslayer optimization of the Gnutella protocol). In Section 2 we analyze and compare the performance (in multi-hop ad hoc networks) of 802.11 card with those of the enhanced card we designed and implemented. First we present a simulation study that shows the effectiveness of our solutions in several scenarios that (in the literature) are known as critical for 802.11 cards. Then we present experimental results obtained in a 4-node networks. In this network the nodes use either the 802.11 card or our enhanced card. Experimental results confirm previous simulation studies. In addition, they point out additional advantages of the enhanced card when used in a real environment with highly variable channel conditions. Section 3 is devoted to analyzing MobileMAN networking protocols that use the one-hop transmission services provided by the network interface card to construct end-to-end (reliable) delivery services. Specifically, we first present our experimental results related to OLSR and AODV in small scale networks with node mobility. These results complete the study reported in Deliverable D8. Secondly, we report the performance results of our mechanism for reliable forwarding which exploits cross layer interactions (REEF). This study has been performed via simulation by exploiting our extension of the NS-2 environment. The section ends presenting MOBILEMAN IST-2001-38113 October 2005 Deliverable D16 4 of 222 experimental results of our transport protocol, TPA. Experimental results confirm the observations obtained via simulation and reported in Deliverable D13. Section 4 is devoted to the interconnection of MobileMAN islands to the Internet. Our solutions have been briefly described in Deliverable D14 where we presented the software developed to support the interconnection. For completeness in this deliverable we first present a refined description of our solution and then we report the experimental results that confirm the effectiveness of our approach. Section 5 addresses the enforcement of cooperation within a MANET. Specifically, the section presents an in depth analysis of CORE, i.e., our mechanism to address cooperation issues. The features of CORE are analyzed in terms of simulation metrics that we deem relevant to assess the basic properties of a cooperation enforcement mechanism: the energetic cost beard by CORE-enabled nodes and the efficiency of the detection and punishment mechanisms used in CORE. Simulation results are used to understand if and when a mechanism to distribute reputation information could be necessary in order to improve punishment efficiency: reputation distribution is an optional feature of the CORE mechanism and constitutes to discriminate between CORE and other reputation-based cooperation enforcement mechanisms. Section 6 deals with the MobileMAN middleware platforms. Performance studies are used to show the effectiveness of the cross-layer optimizations. Specifically, we considered two well-known p2p platforms, Gnutella and Pastry, which represent un-structured and structured overlays, respectively. In the case of Gnutella the study has been performed via simulation; while for Pastry, which is part of the MobileMAN architecture, we performed an experimental study. Specifically, in the Pastry case, we present a set of experimental results obtained by comparing in a small testbed the performance of Pastry with those of CrossROAD. CrossROAD is our proposal to enhance Pastry by exploiting cross-layer interactions. Both Gnutella and Pastry studies clearly pointed out that cross-layer optimization are mandatory to achieve good performance in a mobile ad hoc network. In Section 7 we investigate the quality of service experienced by the three applications we selected to test the MobileMAN architecture: UDDI, a whiteboard application (WB) and a VoIP session. In all cases we evaluated the application performance when running on top of a small MobileMAN testbed. In the case of UDDI and WB, we tested both the legacy and cross-layer architecture. Section 8 concludes the deliverable by investigating the behavior and performance of a medium-scale MobileMAN network made of (up to) 23 nodes. MOBILEMAN IST-2001-38113 October 2005 Deliverable D16 5 of 222 Table of