Cambridge

Peer-to-peer technologies are increasingly becoming the medium of choice for delivering media content, both professional and homegrown, to large user populations. Indeed, current P2P swarming systems have been shown to be very efficient for large-scale content distribution with few server resources. However, such systems have been designed for generic file distribution and provide a limited user experience for viewing media content. For example, users need to wait to download the full video before they can start watching it. In general, the main challenge resides in designing systems that ensure that users can start watching a movie at any point in time, with small start-up times and sustainable playback rates.

Digital media companies have recently started embracing P2P networks as an alternative distribution channel. However, with current P2P swarming systems users need to download the full video, and hence, wait a long time before they can start watching it. While a lot of effort has gone into optimizing the distribution of large files, little research has been done on how to enable Video-on-Demand (VoD) functionality with P2P swarming systems. The main challenges reside in ensuring that users can start watching a movie at any point in time, with small start-up times and sustainable playback rates.

This paper aims to quantify two common assumptions about social tagging: (1) that tags are "meaningful" and (2) that the tagging process is influenced by tag suggestions. For (1), we analyze the semantic properties of tags and the relationship between the tags and the content of the tagged page. Our analysis is based on a corpus of search keywords, contents, titles, and tags applied to several thousand popular Web pages. Among other results, we find that the more popular tags of a page tend to be the more meaningful ones. For , we develop a model of how the influence of tag suggestions can be measured. From a user study with over 4,000 participants, we conclude that roughly one third of the tag applications may be induced by the suggestions. Our results would be of interest for designers of social tagging systems and are a step towards understanding how to best leverage social tags for applications such as search and information extraction.

There has been extensive work on network architectures that support multi-path routing to improve performance in wireless mesh networks. However, previous work uses ad-hoc design principles that cannot guarantee any network-wide performance objectives such as conjointly maximizing resource utilization and improving fairness. In parallel, numerous theoretical results have addressed the issue of optimizing a combined metric of network utilization and fairness using techniques based on back-pressure scheduling, routing and flow control. However, the proposed theoretical algorithms are extremely difficult to implement in practice, especially in the presence of 802.11 MAC and TCP.

Round trip delays constitute the basic network measure that can be obtained by end systems without any network support. Our aim is to design measurement-based admission control strategies for streaming applications based on such minimal feedback on network state. To this end we discuss simple statistical models of packet round trip delays accross either local or wide area networks. We observe that the delay component due to queueing scales like the reciprocal of the spare capacity, at least in a 'heavy traffic' regime where spare capacity is scarce. Our models also allow to capture the correlations between consecutive measurements. Based on these results we propose a two-stage strategy for inferring spare capacity along a network path. We show consistency of this estimate, and analyse its asymptotic variance when the number of samples becomes large. We have experimented these strategies in a local network environment. We observe a good match between theory and practice for switched Ethernets. Surprisingly, the match deteriorates only slightly when the network path comprises hubs, although the theoretical models seem to be less applicable to such technology.

The current Internet features traffic from diverse applications; ranging from delay-sensitive web browsing to delay-insensitive data file transfers. This motivates service differentiation, yet router-centric solutions, e.g. diffserv, have not been widely deployed. The current practice relies on a limited service differentiation at network edges (e.g. through traffic managing middle-boxes or by the end-hosts). End-hosts often implement such emulators of low priority service to differentiate low and normal priority traffic. A low priority service may fit well for some applications, e.g., software updates, but may not be adequate for bulk file transfers that aim at large throughputs. A scenario motivating home users is that of many simultaneous bulk transfers. This is a common feature of peer-to-peer file sharing applications.

We consider collaborative tagging systems where users can attach tags to information objects. Such systems are widely used to add keywords meta-data to photos, videos, or web pages (social bookmarking applications). The meta-data is then used by information retrieval mechanism to provide accurate query answers. To that end, the goal of collaborative tagging systems is to quickly discover the true ranking of a tag for an information object with respect to a given ranking criteria. In this paper, we consider the popularity rank as a ranking criteria. Many collaborative tagging systems help users tagging of an object by making suggestions based on the tagging history of an information object. The problem with making tag suggestions is that they may reinforce some tags and the system fails to discover the true popularity rank of a tag for a given information object. To investigate this issue, we propose and study several algorithms for ranking and suggesting tags in collaborative tagging systems where we focus on the following design objectives: (a) learn true popularity rank, (b) make relevant suggestions, and (c) learn fast. We find that simple incremental updates of the suggestion set, which besides suggestion set size require no configuration constants, offer good performance. Performance is evaluated by analysis and numerical results for which we used a dataset of complete tagging histories of urls that we crawled from a popular social bookmarking web service over a month period.

Handling interference is one of the major challenges in the design of multi-user distributed wireless systems. In current systems, interference is managed through carrier sensing mechanisms such as CSMA/CA and through MAC algorithms based on random back-off. However, the asymmetry in channel sensing inevitably causes degraded throughput and fairness issues, such as those caused by hidden terminal problems. We propose Con-traFlow, a solution based on self-interference cancellation and innovative scheduling mechanisms that increases spatial reuse, eliminates hidden terminals, and rectifies decentralized coordination inefficiencies among nodes, thereby improving fairness. Self-interference cancellation is a technique allowing a node to cancel its own transmitted signal and hence to successfully receive data while transmitting on the same channel. We demonstrate the feasibility of such techniques in a low power WPAN setting, using Lyrtech software-defined radios. Self-interference cancellation repairs carrier sensing, making it possible to successfully eliminate hidden terminal problems, even when using current multi-user MAC protocols; but it also provides the opportunity to design new distributed MAC scheduling algorithms that increase the spatial reuse and solve most of the fairness problems associated with current algorithms. We use simulations to illustrate the performance gains achieved when ContraFlow is used and we obtain both a throughput increase over current systems, as well as a significant improvement in fairness.

We consider the problem of ranking and suggesting popular items based on user feedback that appears in applications such as social tagging and search query suggestions. In particular, we assume that the user feedback is generated as follows. The system suggests to each user a (small) subset of items from the set of all possible items. The user can then choose an item from her suggestion set, or alternatively choose an item from the set of all possible items. Using this feedback, the goal is to quickly learn the true popularity of items, and hence being able to suggest items to users that are indeed popular. The difficulty that arises in this context is that making suggestions to users can reinforce the popularity of some items, and hence distort the resulting item ranking. In this paper, we provide an analysis of this problem. We first formally show that suggesting items to users can indeed lead to a skewed popularity ranking of items. We then propose several algorithms for ranking and suggesting items, and study their performance. In addition, we illustrate our results using a numerical cased study that is based on the inferred popularity of tags from a month-long crawl of a popular social bookmarking service. While "naïve" algorithms can lead to a skewed ranking, our results suggests that there exist simple algorithms for ranking and suggesting items that lead to good performance.

Compared to SAT, there is no simple concept of what a solution to a QBF problem is. Furthermore, as the series of QBF evaluations shows, the QBF solvers that are available often disagree. Thus, proof generation for QBF seems to be even more important than for SAT. In this paper we propose a new uniform proof format, which captures refutations and witnesses for a variety of QBF solvers, and is based on a novel extended resolution rule for QBF. Our experiments show the flexibility of this new format. We also identify shortcomings of our format and conjecture that a purely resolution based proof calculus is not powerful enough to trace the most efficient solvers.

Existing program analysis tools that implement abstraction rely on saturating procedures to compute over-approximations of fixpoints. As an alternative, we propose a new algorithm to compute an over-approximation of the set of reachable states of a program by replacing loops in the control flow graph by their abstract transformer. Our technique is able to generate diagnostic information in case of property violations, which we call leaping counterexamples. We have implemented this technique and report experimental results on a set of large ANSI-C programs using abstract domains that focus on properties related to string-buffers.

With the availability of multi-core processors and large-scale computing clusters, the study of parallel algorithms has been revived throughout the industry. We present a portfolio approach to deciding the satisfiability of SMT formulas, based on the recent success of related algorithms for the SAT problem. Our parallel version of Z3 outperforms the sequential solver, with speedups of well over an order of magnitude on many benchmarks.

Modern termination provers rely on a safety checker to construct disjunctively well-founded transition invariants. This safety check is known to be the bottleneck of the procedure. We present an alternative algorithm that uses a light-weight check based on transitivity of ranking relations to prove program termination. We provide an experimental evaluation over a set of 87 Windows drivers, and demonstrate that our algorithm is often able to conclude termination by examining only a small fraction of the program. As a consequence, our algorithm is able to outperform known approaches by multiple orders of magnitude.

In recent years, bit-precise reasoning has gained importance in hardware and software verification. Of renewed interest is the use of symbolic reasoning for synthesising loop invariants, ranking functions, or whole program fragments and hardware circuits. Solvers for the quantifier-free fragment of bit-vector logic exist and often rely on SAT solvers for efficiency. However, many techniques require quantifiers in bit-vector formulas to avoid an exponential blow-up during construction. Solvers for quantified formulas usually flatten the input to obtain a quantified Boolean formula, losing much of the word-level information in the formula. We present a new approach based on a set of effective word-level simplifications that are traditionally employed in automated theorem proving, heuristic quantifier instantiation methods used in SMT solvers, and model finding techniques based on skeletons/templates. Experimental results on two different types of benchmarks indicate that our method outperforms the traditional flattening approach by multiple orders of magnitude of runtime.

Practical software verification is dominated by two major classes of techniques. The first is model checking, which provides total precision, but suffers from the state space explosion problem. The second is abstract interpretation, which is usually much less demanding, but often returns a high number of false positives. We present LOOPFROG, a static analyzer that combines the best of both worlds: the precision of model checking and the performance of abstract interpretation. In contrast to traditional static analyzers, it also provides 'leaping' counterexamples to aid in the diagnosis of errors.

We present a technique for program termination analysis based on loop summarization. The algorithm relies on a library of abstract domains to discover well-founded transition invariants. In contrast to state-of-the-art methods it aims to construct a complete ranking argument for all paths through a loop at once, thus avoiding expensive enumeration of individual paths. Compositionality is used as a completeness criterion for the discovered transition invariants. The practical efficiency of the approach is evaluated using a set of Windows device drivers.

The increasing popularity of automated tools for software and hardware verification puts ever increasing demands on the underlying decision procedures. This paper presents a framework for distributed decision procedures (for first-order problems) based on Craig interpolation. Formulas are distributed in a lazy fashion, i.e., without the use of costly decomposition algorithms. Potential models which are shown to be incorrect are reconciled through the use of Craig interpolants. Experimental results on challenging propositional satisfiability problems indicate that our method is able to outperform traditional solving techniques even without the use of additional resources.

Recent advances in software termination analysis have shown that program termination can be decided efficiently for many practically relevant problems, despite the fact that the Halting Problem in general is undecidable. This dissertation presents a new algorithm for termination analysis, called Compositional Termination Analysis, which is based on compositional (or transitive) transition invariants. This algorithm depends on an underlying ranking relation synthesis engine and this dissertation presents two such engines that are able to synthesize Bit-Vector ranking relations. This class of ranking relations is especially important for verification of embedded software or for software that interacts with hardware, like device drivers.