Raquel Morera

An analytical approach to the performance evaluation of mobility protocols: the handoff delay case

Under few assumptions (random walk mobility model, transmission cost proportional to distance, minimum distance routing), this paper describes closed form expressions to quantify the handoff delay of two mobility mechanisms, the home agent-based and the binding update-based ones. This allows us to estimate the performance of different protocols, choose under what conditions each protocol performs best and better tune the protocol parameters to increase performance. The comparative analysis here proposed allows us to quantify the benefits and tradeoffs of each of the approaches and take appropriate decisions when different mobility protocols are at our choice. This is an important result in many applications, as it gives a first order approximation on whether to use the home agent option or route optimization option in MIPv6 as a function of the distance of the mobile entities. Although we have focused in this paper basically on two mobility managements mechanisms, the analysis is easily extensible to other approaches.

Adapting DNS to dynamic ad hoc networks

The limitations in the domain naming service (DNS), such as vulnerability to denial of service attacks, configuration errors, and loss of root, home or peer name servers, have lead to proposals for more robust, secure, flexible, and dynamic alternatives. The ad hoc bandwidth constrained nature of future military networks, such as FCS and WIN-T, together with the secure overlay networks created by HAIPE IPsec tunnels, exacerbates the limitations in DNS. This paper proposes architectural guidelines and a new autoconfiguration protocol to fix some key limitations in applying current DNS to future military networks. This approach requires no change to standard domain name space, and no changes to existing DNS client or name server software. The most radical change is to replace the static DNS roles and linkages with automatically configured ones. Moreover, the configurations will be dynamically modified based on the current network environment and mission needs. In addition to the obvious benefits of removing crippling configuration errors, we show that good dynamic DNS configurations can greatly improve robustness and efficiency after loss of nodes or changing link bandwidths

Robust router reconfiguration in large dynamic networks

In the current IP protocol suite there is a tradeoff between scalability and the maximum allowable mobility. Flat MANET routing protocols can deal with high mobility, but cannot scale to support the thousands of mobile routers needed in the Armys Future Combat System. Using IP subnets and hierarchical domains provides scalability, but breaks existing configuration protocols (A. McAuley, et al., 2002) in high mobile networks (e.g., network splits lead to conflicting routing tables). This evidences the need for router mobility detection and dynamic reconfiguration mechanisms in scalable mobile networking. This paper describes a new beacon protocol to allow rapid detection of IP router movements, including network splitting and merging, within large hierarchical domains. It detects movement by identifying each IP subnets and higher-level domains with a beacon node; and, by not relying on network configuration, IP routing, or any fixed nodes, it is robust to any network changes. Nodes can optionally report topology changes to the beacon node with low overhead.

An analytical approach to the performance evaluation of mobility protocols: the overall mobility cost case

Although there are a plethora of mobility management solutions proposed in the literature, there is no method for analytical comparison that helps identify optimal solutions in a specific scenario. Since simulations and implementations can provide only limited information, we propose an analytical approach for the comparative analysis of mobility protocols. By decomposing existing protocols into their building blocks, we obtain general cost functions for different mobility mechanisms. Our analysis allows us to identify network and topology conditions under which a certain protocol performs better than another, according to specific criteria. We present results for the signaling and overhead case, whereas the handoff delay case is addressed in S. Galli et al. (2004).

Modeling topology dissemination for routing in future force networks

The network-centric future force must support a large and diverse group of communication nodes. While much network design in networks such as future combat systems (FCS) and warfighter information network- tactical (WIN-T) can be done using existing TCP/IP protocols, there are important performance limits when the network conditions become more extreme (e.g., highly mobile or dense). A solution for a particular environment requires performance approximation over a wide range of conditions and detailed simulation. This paper provides models and analysis of the expected user performance of different approaches to flooding link state routing information in wireless ad hoc networks for routing update. We compare the use of flat flooding, multi-point relays (MPRs) and connected dominating sets (CDSs), assuming other parts of the routing protocol are taken from standard link state routing protocol (e.g. OLSR) or open shortest path first protocol (OSPF), widely used in the Internet and WIN-T to disseminate routing updates. In particular, we investigate what conditions will force the flat link state routing update mechanisms in OSPF to be augmented by more scalable solutions, such as OLSR MPRs and OSPF-MANET CDSs. Although the existing literature provides a variety of models for flat flooding and MPRs, it lacks similar analytical work for relays placement under the CDS approach. The only results available are obtained by simulation. In an attempt to compare all approaches under a similar analytical framework we focused on creating analytic models for CDS relay placements and deriving closed analytical formulae For this paper, we select one such model - the Hexagon - to represent the CDS approach in the comparative analysis, as it provides the lowest routing overhead among other properties. We also briefly introduce our heuristic that simulates the latter model as close as possible. Analysis shows the difference in the impact of conditions on key performance metrics. A key understanding of the models is the impact of network density D on routing overhead. As D increases, existing asymptotic results show that the overhead associated with the MPR approach (0(D5/3)) is much less than this associated with flat flooding (0(D3)); however, our new models show that the overhead of the CDS approach is asymptotically the lowest (0(D)). This improved scalability could be critical to dense deployments of future military networks. The models also designate a tradeoff between routing overhead and routing stretch. In particular, the CDS approach has an asymptotic worst case stretch of at most 1.33 times more than other schemes. Other results show that the CDS approach simplifies MAC scheduling and can increase network lifetime. The models should thus enable network designers to understand when OSPF is adequate for future military networks, and when network conditions dictate consideration of topology update mechanisms such as these of OLSR MPRs or CDSs.

Fault localization and self-healing with dynamic domain configuration

Compared to their commercial counterparts, future battlefield networks require much more extensive fault management and automation mechanisms. Much work has been done to improve these functions, since survivability and automation are seen as critical to the Armys next generation tactical and strategic battlefield networks such as FCS; however, they are generally treated separately. This paper highlights the synergy between these functions. In particular we show that by reconfiguring domains, as fault localization and multilayer self-healing algorithms mandate, we can help speed the process in isolating the cause of many of these alarms and help solve some of the network problems.

Name and address decoupling in support of dynamic networks

Many of the problems with dynamic networks, such as those envisioned in Future Combat Systems, lie in the fundamental IP architecture. A well-known problem is the coupling of names with the IP addresses used for routing. Names can reflect topology (e.g., domain names reflect their topological domain), which can limit survivability. More important, IP addresses identify hosts in the transport and application layers, which limits their ability to support multi-homing, allow Network Address Translation and rapidly reflect topological changes. Ideally, names should identify entities within a purely logical structure, with no correlation to topology, while addresses dynamically reflect the global network topology. This paper describes our approach to decoupling names and addresses based on some proposed changes to the current TCP/IP protocol suite. First, creating a Session Maintenance Protocol (SMP) with a new local Session Identifier that supports transparent changes to IP addresses. Second, replacing DNS, which simply returns the IP address associated with a name, with a Logic Name Server (LNS) that routes messages to the Location Server associated with a name. Third, identify small changes to many transport and higher layer protocols required to remove any dependence on IP addresses. Finally, we compare our approach to existing solutions in support of dynamic networks, such as Mobile IP and HIP.

Flexible autoconfigured domains for more scalable, efficient and robust battlefield networks

Splitting a network into independent domains can significantly improve network performance, for functions such as routing, configuration and security. Limiting domain size improves protocol scalability, survivability, and efficiency, particularly in more dynamic networks such as that envisioned in future battlefield networks. This paper describes our general framework for building autoconfigurable, hierarchical domains and how it can be applied to different types of networks: from simple IP and adhoc subnets, to large collections of nodes connected by bridges or routers. We also show how the domains can be split and merged as nodes and even whole networks move.

A secure virtual point of service for purchasing digital media content over 3G wireless networks

We propose the notion of a Secure Virtual Point of Service (SVPOS) as a network-centric transaction server that facilitates the enhancement of 3G cell phones with a ‘mobile wallet’ capability allowing 3G subscribers to use their cell phones (or, in fact, other preferred mobile gadgets) for their daily transactions and payments. This paper shows how to design an SVPOS and an associated operator/subscriber/merchant protocol for purchasing digital media content over 3G networks. The resulting protocol guarantees a number of desirable privacy and security properties, such as privacy/anonymity of 3G subscribers (i.e. no identities, credit information or credit card numbers are revealed by subscriber to merchants), and protection of 3G operator, merchant, subscriber against various types of malicious behaviour, including transaction repudiation. The proposed SVPOS is ‘built’ on top of the Hypertext Transfer Protocol (HTTP), utilizes the 3rd Generation Partnership Project (3GPP) Generic Authentication Architecture (GAA) for subscriber and merchant authentication, and has an implicit key distribution mechanism that easily provides necessary encryption keys for the novel non-repudiation mechanism of SVPOS. It also sends the necessary records of the transactions to the accounting entity of the network for charging and billing through standard protocols (e.g. Parlay-X). Copyright

Impact of channel errors in proactive routing protocols

In proactive routing protocols (such as OLSR) nodes flood their neighborhood information to the entire network. Every node then constructs a topology map which represents its view of the network. This map is taken as a basis to compute the “best” path to the destination. Inaccurate topology information may result in one of the following: a) packets may be sent in a path that no longer exists; b) routing loops may be formed when two neighboring nodes have a different view of the network; c) packets may be sent through a longer path. In the latter, packets are still being delivered to the destination, thus inaccurate topology information does not necessarily result in packet delivery failures but in this case in increased stretch. In wireless Mobile Ad Hoc Networks, channel errors originated by poor channel quality or mobility can cause the loss of topology control packets. This may consequently result in nodes having an erroneous and/or inconsistent view of the network topology. In this paper, we first describe the Component Based Routing (CBR) approach to routing protocol design and then focus on the evaluation of the topology dissemination component. We analyze the impact that channel errors and node mobility have on the successful delivery of topology control packets and routing protocol performance. We derive models that allow us evaluate performance for different network sizes and node densities. We evaluate the tradeoffs between accuracy of topology information at any given time, power consumption and overhead for both optimized flooding mechanisms (such as those that are used in OLSR) and non-optimized flooding mechanisms (e.g. flat flooding).