Shyam Kapadia

PAVAN: a policy framework for content availabilty in vehicular ad-hoc networks

Advances in wireless communication, storage and processing are realizing next-generation in-vehicle entertainment systems. Even if hundreds of different video or audio titles are stored among several vehicles in an area, only a subset of these titles might be available to a given vehicle depending on its current location, intended path, and the dynamics of its ad-hoc network connectivity. The vehicles entertainment system must somehow predictively determine which titles are available either immediately or within the future d time units, so that the user can select a title to view. The available title list must seek to satisfy the user by striking a delicate balance between showing far fewer titles than can actually be accessed and showing too many titles that cannot be accessed. In addition to defining this availability problem, we make two key contributions. First, a two-tier system architecture which leverages the low-rate cellular infrastructure as a control network for the high-rate data network consisting of the vehicular ad-hoc network. Second, PAVAN as a policy framework for predicting the availability of a title. We describe several variants of PAVAN which incorporate information based on a Markov mobility model, spatio-temporal look-ahead, and title replications. Our results demonstrate that the quality of PAVANs predictions is critically dependent on title degree of replication, as well as its relative size with respect to the trip duration. When degree of replication is below a certain threshold, PAVAN with content density information and the predictive mobility model is shown to provide the best overall performance.

Evaluation of 802.11a for streaming data in ad-hoc networks

Advances in communication and processing have made ad-hoc networks of wireless devices a reality. One application is home entertainment systems where multiple Home-to-Home (H20) devices collaborate as peers to stream audio and video clips to a household. In this study, we investigate the feasibility of IEEE 802.11a protocol in combination with both TCP and UDP to realize a H20 device. Challenges include lossy connections, unfair allocation of bandwidth between multiple simultaneous transmissions, and the exposed node limitation [22], [19], [13], [4]. Our primary contribution is an empirical study of 802.11a to quantify these factors and their significance. Our multi-dimensional experimental design consists of the following axes: distance between participating devices, number of intermediate H20 devices used to route a stream from a producing H20 device to a consuming H2O device, and simultaneous number of active streams in the same radio range. Both operating system and application level routing were considered_ Obtained results demonstrate the following lessons. First, with a multi-hop UDP transmission, in the absence of congestion control, transient bottlenecks result in a high loss rate. Hence, a transport protocol with congestion control is essential for streaming of continuous media within a H2O cloud. Second, 802.11a does not drop TCP connections in the presence of many competing transmissions (802.11b drops connections [22]). Third, we observed fairness when transmitting several hundred Megabytes (MB) of data, among multiple competing 1- hop TCP and UDP flows. Fourth, while there is unfair allocation of bandwidth with an exposed node, the observed bandwidths are sufficient to stream a high-quality video clip (with a 4 Mbps display bandwidth requirement). These results indicate streaming of data is feasible with an ad-hoc network of wireless devices employing the 80211a protocol.

Comparative analysis of push-pull query strategies for wireless sensor networks

We present a comparative mathematical analysis of two important distinct approaches to hybrid push-pull querying in wireless sensor networks: structured hash-based data-centric storage (DCS) and the unstructured comb-needle (CN) rendezvous mechanism. Our analysis yields several interesting insights. For ALL-type queries pertaining to information about all events corresponding to a given attribute, we examine the conditions under which the two approaches outperform each other in terms of the average query and event rates. For the case of ANY-type queries where it is sufficient to obtain information from any one of the desired events for a given attribute, we propose and analyze a modified sequential comb-needle technique (SCN) to compare with DCS. We find that DCS generally performs better than CN/SCN for high query rates and low event rates, while CN/SCN perform better for high event rates. Surprisingly, for the cases of ALL-type aggregated queries and ANY-type queries, we find that there exist “magic number” event rate thresholds, independent of network size or query probability, which dictate the choice of querying protocol. While our analysis is based on a single-sink square-grid deployment, we believe the insights can be generalized to random deployments.

Empirical evaluation of querying mechanisms for unstructured wireless sensor networks

In the last few years, several studies have analyzed the performance of flooding and random walks as querying mechanisms for unstructured wireless sensor networks. However, most of the work is theoretical in nature and while providing insights into the asymptotic behavior of these querying mechanisms, does not account for the non-idealities faced by the network in real deployments. In this paper, we propose a 3-way handshake protocol as a reliable implementation of a random walk and compare its performance with flooding in real environments. The metrics considered are delay, reliability and transmission cost. Our initial results suggest that flooding is better suited for low-interference environments, while random walks might be a better option in networks with high interference. We also present possible research directions to improve the performance oflooding and random walks.

Data Delivery in Delay Tolerant Networks: A Survey

Delay-Tolerant Networks (Fall (2003)), also called disruption tolerant networks (DTNs), represent a fairly new networking paradigm that allows inter-connection between devices that current networking technology cannot provide. There are a wide variety of networks where an end-to-end connection between a given source and destination may never be present. Consequently, traditional routing protocols cannot be directly applied in these scenarios for delivering data. However, if one were to take the graph formed by the nodes based on their connectivity dictated by their radio range and consider the overlap not only over space but also time then there is a high likelihood that the network will appear as a single connected component. So while at any given instant, the network may not be connected, it may still be possible to route data from a source to a destination. DTNs are sometimes also called Intermittently-Connected Mobile Networks (ICMNs). The primary goal in such networks is to get the information from a source to the destination; these networks can tolerate a relatively higher delay. A wide variety of ”challenged” networks fall under this category ranging from outer-space networks, under-water networks, wireless sensor networks, vehicular networks, sparse mobile ad-hoc networks etc. Students moving about in a college campus (Hsu & Helmy (2006)), or buses moving about in a small metropolitan area (Burgess et al. (2006)), or a wireless sensor network with some mobile nodes (Shah et al. (2003); Juang et al. (2002)) acting as relays to assist in the data-collection phase provide representative examples of DTNs. This chapter strives to provide a survey of some of the most relevant studies that have appeared in the domain of data delivery in delay tolerant networks. First, we introduce some fundamental challenges that are unique to DTNs. Then we present the major parameters of interest that various proposed routing solutions have considered, examples include end-to-end delay, throughput, mobility model of the nodes, energy efficiency, storage etc. Subsequently, we provide a classification of various approaches to routing in DTNs and pigeon-hole the major studies that have appeared in the last few years into the classified categories.

Static Replication Strategies for Content Availability in Vehicular Ad-hoc Networks

This study investigates replication strategies for reducing latency to desired content in a vehicular peer-to-peer network. We provide a general constrained optimization formulation for efficient replication and study it via analysis and simulations employing a discrete random walk mobility model for the vehicles. Our solution space comprises of a family of popularity based replication schemes each characterized by an exponent n. We find that the optimal replication exponent depends significantly on factors such as the total system storage, data item size, and vehicle trip duration. With small data items and long client trip durations, n ∼ 0.5 i.e., a square-root replication scheme provides the lowest aggregate latency across all data item requests. However, for short trip durations, n moves toward 1, making a linear replication scheme more favorable. For larger data items and long client trip durations, we find that the optimal replication exponent is below 0.5. Finally, for these larger data items, if the client trip duration is short, the optimal replication exponent is found to be a function of the total storage in the system. Subsequently, the above observations are validated with two real data sets: one based on a city map with freeway traffic information and the other employing encounter traces from a bus network.

An evaluation of availability latency in carrier-based wehicular ad-hoc networks

On-demand delivery of audio and video clips in peer-to-peer vehicular ad-hoc networks is an emerging area of research. Our target environment uses data carriers, termed zebroids, where a mobile device carries a data item on behalf of a server to a client thereby minimizing its availability latency. In this study, we quantify the variation in availability latency with zebroids as a function of a rich set of parameters such as car density, storage per device, repository size, and replacement policies employed by zebroids. Using analysis and extensive simulations, we gain novel insights into the design of carrier-based systems. Significant improvements in latency can be obtained with zebroids at the cost of a minimal overhead. These improvements occur even in scenarios with lower accuracy in the predictions of the car routes. Two particularly surprising findings are: (1) a naive random replacement policy employed by the zebroids shows competitive performance, and (2) latency improvements obtained with a simplified instantiation of zebroids are found to be robust to changes in the popularity distribution of the data items.

Comparison of replication strategies for content availability in C2P2 networks

This study investigates alternative continuous media replication techniques and their impact on content availability in a mobile car-to-car peer-to-peer (C2P2) network of devices. Using aggregate availability latency as a metric, we compare a simple random replication mechanism with a family of techniques that compute the degree of replication for each title based on its popularity, i.e., frequency of access. We use a simulation study along with some supporting analytical analysis for this comparison. Obtained results demonstrate the following key lesson. When total storage capacity of the network is significantly larger than the clip repository size, a random replication technique is sufficient. Otherwise, there is a large parameter space where the frequency-based replication schemes provide superior performance.

Infection spread in wireless networks with random and adversarial node mobilities

We study the process of the spread of an infection among mobile nodes moving on a finite, grid based map. A random walk and a novel adversarial model are considered as two extreme cases of node mobility. With <i>N</i> nodes, we present analytical and simulation results for both mobility models for a square grid map with size √<i>G</i> × √<i>G</i>. A key finding is that with random mobility the total time to infect all nodes decreases with <i>N</i> while with an adversarial model we observe a reverse trend. Specifically, the random case results in a total infection time of Θ(<i>G</i>log<i>G</i>log<i>N</i>/(<i>N</i>) as opposed to the adversarial case where the total infection time is found to be Θ(√(<i>G</i>log(<i>N</i>). We also explore the possibility of emulating such an infection process as a mobile interaction game with wireless sensor motes, and the above results are complimented by traces obtained from an empirical study with humans as players in an outdoor field.

An evaluation of location-demographic replacement policies for zebroids

In an ad-hoc network of mobile devices, a device is termed a zebroid when it carries a data item residing on a server-device to a client-device referencing that item. The system employs a zebroid when its travel path intersects that of the server and the client and this information is known in advance. The motivation for use of a zebroid is to minimize the delay incurred by the client for the referenced data item, termed the availability latency. This paper considers the performance of alternative policies that manage the identity of data items assigned to a zebroid when its storage is exhausted. One novel policy is LoDeR that manages storage of zebroids based on the demographics of a geographical region where the zebroid rendezvous with the client. Experimental results demonstrate a host of tradeoffs between the different performance metrics such as the number of data items lost by a policy and the percentage of requests that reference these data items, availability latency, and number of replaced data items. The mobility model has a significant impact on these metrics for a given policy.