Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing

The proliferation of inexpensive workstations and networks has prompted several researchers to use such distributed systems for parallel computing. Attempts have been made to offer a shared-memory programming model on such distributed memory computers. Most systems provide a shared-memory that is coherent in that all processes that use it agree on the order of all memory events. This dissertation explores the possibility of a significant improvement in the performance of some applications when they use non-coherent memory. First, a new formal model to describe existing non-coherent memories is developed. I use this model to prove that certain problems can be solved using asynchronous iterative algorithms on shared-memory in which the coherence constraints are substantially relaxed. In the course of the development of the model I discovered a new type of non-coherent behavior called Local Consistency. Second, a programming model, Mermera, is proposed. It provides programmers with a choice of hierarchically related non-coherent behaviors along with one coherent behavior. Thus, one can trade-off the ease of programming with coherent memory for improved performance with non-coherent memory. As an example, I present a program to solve a linear system of equations using an asynchronous iterative algorithm. This program uses all the behaviors offered by Mermera. Third, I describe the implementation of Mermera on a BBN Butterfly TC2000 and on a network of workstations. The performance of a version of the equation solving program that uses all the behaviors of Mermera is compared with that of a version that uses coherent behavior only. For a system of 1000 equations the former exhibits at least a 5-fold improvement in convergence time over the latter. The version using coherent behavior only does not benefit from employing more than one workstation to solve the problem while the program using non-coherent behavior continues to achieve improved performance as the number of workstations is increased from 1 to 6. This measurement corroborates our belief that non-coherent shared memory can be a performance boon for some applications.

An Implementation of Mermera: A Shared Memory System that Mixes Coherence with Non-coherence

Coherent shared memory is a convenient, but inefficient, method of inter-process communication for parallel programs. By contrast, message passing can be less convenient, but more efficient. To get the benefits of both models, several non-coherent memory behaviors have recently been proposed in the literature. We present an implementation of Mermera, a shared memory system that supports both coherent and non-coherent behaviors in a manner that enables programmers to mix multiple behaviors in the same program[HS93]. A programmer can debug a Mermera program using coherent memory, and then improve its performance by selectively reducing the level of coherence in the parts that are critical to performance. Mermera permits a trade-off of coherence for performance. We analyze this trade-off through measurements of our implementation, and by an example that illustrates the style of programming needed to exploit non-coherence. We find that, even on a small network of workstations, the performance advantage of non-coherence is compelling. Raw non-coherent memory operations perform 20-40~times better than non-coherent memory operations. An example application program is shown to run 5-11~times faster when permitted to exploit non-coherence. We conclude by commenting on our use of the Isis Toolkit of multicast protocols in implementing Mermera.

New Notions of Reduction and Non-Semantic Proofs of β-Strong Normalization in Typed λ-Calculi

Two new notions of reduction for terms of the λ-calculus are introduced and the question of whether a λ-term is beta-strongly normalizing is reduced to the question of whether a λ-term is merely normalizing under one of the new notions of reduction. This leads to a new way to prove beta-strong normalization for typed λ-calculi. Instead of the usual semantic proof style based on Girard's "candidats de réductibilité'', termination can be proved using a decreasing metric over a well-founded ordering in a style more common in the field of term rewriting. This new proof method is applied to the simply-typed λ-calculus and the system of intersection types.

On the Efficiency and Fairness of Transmission Control Loops: A Case for Exogenous Losses

We postulate that exogenous losses-which are typically regarded as introducing undesirable "noise" that needs to be filtered out or hidden from end points-can be surprisingly beneficial. In this paper we evaluate the effects of exogenous losses on transmission control loops, focusing primarily on efficiency and convergence to fairness properties. By analytically capturing the effects of exogenous losses, we are able to characterize the transient behavior of TCP. Our numerical results suggest that "noise" resulting from exogenous losses should not be filtered out blindly, and that a careful examination of the parameter space leads to better strategies regarding the treatment of exogenous losses inside the network. Specifically, we show that while low levels of exogenous losses do help connections converge to their fair share, higher levels of losses lead to inefficient network utilization. We draw the line between these two cases by determining whether or not it is advantageous to hide, or more interestingly introduce, exogenous losses. Our proposed approach is based on classifying the effects of exogenous losses into long-term and short-term effects. Such classification informs the extent to which we control exogenous losses, so as to operate in an efficient and fair region. We validate our results through simulations.

BoostMap: A Method for Efficient Approximate Similarity Rankings

This paper introduces BoostMap, a method that can significantly reduce retrieval time in image and video database systems that employ computationally expensive distance measures, metric or non-metric. Database and query objects are embedded into a Euclidean space, in which similarities can be rapidly measured using a weighted Manhattan distance. Embedding construction is formulated as a machine learning task, where AdaBoost is used to combine many simple, 1D embeddings into a multidimensional embedding that preserves a significant amount of the proximity structure in the original space. Performance is evaluated in a hand pose estimation system, and a dynamic gesture recognition system, where the proposed method is used to retrieve approximate nearest neighbors under expensive image and video similarity measures. In both systems, BoostMap significantly increases efficiency, with minimal losses in accuracy. Moreover, the experiments indicate that BoostMap compares favorably with existing embedding methods that have been employed in computer vision and database applications, i.e., FastMap and Bourgain embeddings.

A Bayesian Approach for TCP to Distinguish Congestion from Wireless Losses

(This Technical Report revises TR-BUCS-2003-011) The Transmission Control Protocol (TCP) has been the protocol of choice for many Internet applications requiring reliable connections. The design of TCP has been challenged by the extension of connections over wireless links. In this paper, we investigate a Bayesian approach to infer at the source host the reason of a packet loss, whether congestion or wireless transmission error. Our approach is "mostly" end-to-end since it requires only one long-term average quantity (namely, long-term average packet loss probability over the wireless segment) that may be best obtained with help from the network (e.g. wireless access agent).Specifically, we use Maximum Likelihood Ratio tests to evaluate TCP as a classifier of the type of packet loss. We study the effectiveness of short-term classification of packet errors (congestion vs. wireless), given stationary prior error probabilities and distributions of packet delays conditioned on the type of packet loss (measured over a larger time scale). Using our Bayesian-based approach and extensive simulations, we demonstrate that congestion-induced losses and losses due to wireless transmission errors produce sufficiently different statistics upon which an efficient online error classifier can be built. We introduce a simple queueing model to underline the conditional delay distributions arising from different kinds of packet losses over a heterogeneous wired/wireless path. We show how Hidden Markov Models (HMMs) can be used by a TCP connection to infer efficiently conditional delay distributions. We demonstrate how estimation accuracy is influenced by different proportions of congestion versus wireless losses and penalties on incorrect classification.

Addendum to "New Notions of Reduction and Non-Semantic Proofs of β-Strong Normalization in Typed λ-Calculi"

This is an addendum to our technical report BUCS TR-94-014 of December 19, 1994. It clarifies some statements, adds information on some related research, includes a comparison with research be de Groote, and fixes two minor mistakes in a proof.

Robust Identification of Shared Losses Using End-to-End Unicast Probes

ERRATA: We present corrections to Fact 3 and (as a consequence) to Lemma 1 of BUCS Technical Report BUCS-TR-2000-013 (also published in IEEE INCP'2000)[1]. These corrections result in slight changes to the formulae used for the identifications of shared losses, which we quantify.

Scalability of Multicast Delivery for Non-sequential Streaming Access

To serve asynchronous requests using multicast, two categories of techniques, stream merging and periodic broadcasting have been proposed. For sequential streaming access where requests are uninterrupted from the beginning to the end of an object, these techniques are highly scalable: the required server bandwidth for stream merging grows logarithmically as request arrival rate, and the required server bandwidth for periodic broadcasting varies logarithmically as the inverse of start-up delay. However, sequential access is inappropriate to model partial requests and client interactivity observed in various streaming access workloads. This paper analytically and experimentally studies the scalability of multicast delivery under a non-sequential access model where requests start at random points in the object. We show that the required server bandwidth for any protocols providing immediate service grows at least as the square root of request arrival rate, and the required server bandwidth for any protocols providing delayed service grows linearly with the inverse of start-up delay. We also investigate the impact of limited client receiving bandwidth on scalability. We optimize practical protocols which provide immediate service to non-sequential requests. The protocols utilize limited client receiving bandwidth, and they are near-optimal in that the required server bandwidth is very close to its lower bound.

Amorphous Placement and Informed Diffusion for Timely Monitoring by Autonomous, Resource-Constrained, Mobile Sensors

Personal communication devices are increasingly equipped with sensors for passive monitoring of encounters and surroundings. We envision the emergence of services that enable a community of mobile users carrying such resource-limited devices to query such information at remote locations in the field in which they collectively roam. One approach to implement such a service is directed placement and retrieval (DPR), whereby readings/queries about a specific location are routed to a node responsible for that location. In a mobile, potentially sparse setting, where end-to-end paths are unavailable, DPR is not an attractive solution as it would require the use of delay-tolerant (flooding-based store-carry-forward) routing of both readings and queries, which is inappropriate for applications with data freshness constraints, and which is incompatible with stringent device power/memory constraints. Alternatively, we propose the use of amorphous placement and retrieval (APR), in which routing and field monitoring are integrated through the use of a cache management scheme coupled with an informed exchange of cached samples to diffuse sensory data throughout the network, in such a way that a query answer is likely to be found close to the query origin. We argue that knowledge of the distribution of query targets could be used effectively by an informed cache management policy to maximize the utility of collective storage of all devices. Using a simple analytical model, we show that the use of informed cache management is particularly important when the mobility model results in a non-uniform distribution of users over the field. We present results from extensive simulations which show that in sparsely-connected networks, APR is more cost-effective than DPR, that it provides extra resilience to node failure and packet losses, and that its use of informed cache management yields superior performance.

Non-Uniform Reductions

We study properties of non-uniform reductions and related completeness notions. We strengthen several results of Hitchcock and Pavan and give a trade-off between the amount of advice needed for a reduction and its honesty on NEXP. We construct an oracle relative to which this trade-off is optimal. We show, in a more systematic study of non-uniform reductions, that among other things non-uniformity can be removed at the cost of more queries. In line with Post's program for complexity theory we connect such 'uniformization' properties to the separation of complexity classes.

Simultaneous Learning of Non-Linear Manifold and Dynamical Models for High-Dimensional Time Series

The goal of this work is to learn a parsimonious and informative representation for high-dimensional time series. Conceptually, this comprises two distinct yet tightly coupled tasks: learning a low-dimensional manifold and modeling the dynamical process. These two tasks have a complementary relationship as the temporal constraints provide valuable neighborhood information for dimensionality reduction and conversely, the low-dimensional space allows dynamics to be learnt efficiently. Solving these two tasks simultaneously allows important information to be exchanged mutually. If nonlinear models are required to capture the rich complexity of time series, then the learning problem becomes harder as the nonlinearities in both tasks are coupled. The proposed solution approximates the nonlinear manifold and dynamics using piecewise linear models. The interactions among the linear models are captured in a graphical model. The model structure setup and parameter learning are done using a variational Bayesian approach, which enables automatic Bayesian model structure selection, hence solving the problem of over-fitting. By exploiting the model structure, efficient inference and learning algorithms are obtained without oversimplifying the model of the underlying dynamical process. Evaluation of the proposed framework with competing approaches is conducted in three sets of experiments: dimensionality reduction and reconstruction using synthetic time series, video synthesis using a dynamic texture database, and human motion synthesis, classification and tracking on a benchmark data set. In all experiments, the proposed approach provides superior performance.

Non-Rigid Shape from Image Streams

We present a framework for estimating 3D relative structure (shape) and motion given objects undergoing nonrigid deformation as observed from a fixed camera, under perspective projection. Deforming surfaces are approximated as piece-wise planar, and piece-wise rigid. Robust registration methods allow tracking of corresponding image patches from view to view and recovery of 3D shape despite occlusions, discontinuities, and varying illumination conditions. Many relatively small planar/rigid image patch trackers are scattered throughout the image; resulting estimates of structure and motion at each patch are combined over local neighborhoods via an oriented particle systems formulation. Preliminary experiments have been conducted on real image sequences of deforming objects and on synthetic sequences where ground truth is known.

Robust Identification of Shared Losses Using End-to-End Unicast Probes

Current Internet transport protocols make end-to-end measurements and maintain per-connection state to regulate the use of shared network resources. When two or more such connections share a common endpoint, there is an opportunity to correlate the end-to-end measurements made by these protocols to better diagnose and control the use of shared resources. We develop packet probing techniques to determine whether a pair of connections experience shared congestion. Correct, efficient diagnoses could enable new techniques for aggregate congestion control, QoS admission control, connection scheduling and mirror site selection. Our extensive simulation results demonstrate that the conditional (Bayesian) probing approach we employ provides superior accuracy, converges faster, and tolerates a wider range of network conditions than recently proposed memoryless (Markovian) probing approaches for addressing this opportunity.

Efficient Nearest Neighbor Classification Using a Cascade of Approximate Similarity Measures

Nearest neighbor classification using shape context can yield highly accurate results in a number of recognition problems. Unfortunately, the approach can be too slow for practical applications, and thus approximation strategies are needed to make shape context practical. This paper proposes a method for efficient and accurate nearest neighbor classification in non-Euclidean spaces, such as the space induced by the shape context measure. First, a method is introduced for constructing a Euclidean embedding that is optimized for nearest neighbor classification accuracy. Using that embedding, multiple approximations of the underlying non-Euclidean similarity measure are obtained, at different levels of accuracy and efficiency. The approximations are automatically combined to form a cascade classifier, which applies the slower approximations only to the hardest cases. Unlike typical cascade-of-classifiers approaches, that are applied to binary classification problems, our method constructs a cascade for a multiclass problem. Experiments with a standard shape data set indicate that a two-to-three order of magnitude speed up is gained over the standard shape context classifier, with minimal losses in classification accuracy.