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.

Mapping Parallel Iterative Algorithms onto Workstation Networks

For communication-intensive parallel applications, the maximum degree of concurrency achievable is limited by the communication throughput made available by the network. In previous work [HPS94], we showed experimentally that the performance of certain parallel applications running on a workstation network can be improved significantly if a congestion control protocol is used to enhance network performance. In this paper, we characterize and analyze the communication requirements of a large class of supercomputing applications that fall under the category of fixed-point problems, amenable to solution by parallel iterative methods. This results in a set of interface and architectural features sufficient for the efficient implementation of the applications over a large-scale distributed system. In particular, we propose a direct link between the application and network layer, supporting congestion control actions at both ends. This in turn enhances the system's responsiveness to network congestion, improving performance. Measurements are given showing the efficacy of our scheme to support large-scale parallel computations.

OS Support for Portable Bulk Synchronous Parallel Programs

Predictability -- the ability to foretell that an implementation will not violate a set of specified reliability and timeliness requirements -- is a crucial, highly desirable property of responsive embedded systems. This paper overviews a development methodology for responsive systems, which enhances predictability by eliminating potential hazards resulting from physically-unsound specifications. The backbone of our methodology is the Time-constrained Reactive Automaton (TRA) formalism, which adopts a fundamental notion of space and time that restricts expressiveness in a way that allows the specification of only reactive, spontaneous, and causal computation. Using the TRA model, unrealistic systems – possessing properties such as clairvoyance, caprice, infinite capacity, or perfect timing -- cannot even be specified. We argue that this "ounce of prevention" at the specification level is likely to spare a lot of time and energy in the development cycle of responsive systems -- not to mention the elimination of potential hazards that would have gone, otherwise, unnoticed. The TRA model is presented to system developers through the Cleopatra programming language. Cleopatra features a C-like imperative syntax for the description of computation, which makes it easier to incorporate in applications already using C. It is event-driven, and thus appropriate for embedded process control applications. It is object-oriented and compositional, thus advocating modularity and reusability. Cleopatra is semantically sound; its objects can be transformed, mechanically and unambiguously, into formal TRA automata for verification purposes, which can be pursued using model-checking or theorem proving techniques. Since 1989, an ancestor of Cleopatra has been in use as a specification and simulation language for embedded time-critical robotic processes.

Distributed Parallel Computing in Mermera: Mixing Noncoherent Shared Memories

Programmers of parallel processes that communicate through shared globally distributed data structures (DDS) face a difficult choice. Either they must explicitly program DDS management, by partitioning or replicating it over multiple distributed memory modules, or be content with a high latency coherent (sequentially consistent) memory abstraction that hides the DDS' distribution. We present Mermera, a new formalism and system that enable a smooth spectrum of noncoherent shared memory behaviors to coexist between the above two extremes. Our approach allows us to define known noncoherent memories in a new simple way, to identify new memory behaviors, and to characterize generic mixed-behavior computations. The latter are useful for programming using multiple behaviors that complement each others' advantages. On the practical side, we show that the large class of programs that use asynchronous iterative methods (AIM) can run correctly on slow memory, one of the weakest, and hence most efficient and fault-tolerant, noncoherence conditions. An example AIM program to solve linear equations, is developed to illustrate: (1) the need for concurrently mixing memory behaviors, and, (2) the performance gains attainable via noncoherence. Other program classes tolerate weak memory consistency by synchronizing in such a way as to yield executions indistinguishable from coherent ones. AIM computations on noncoherent memory yield noncoherent, yet correct, computations. We report performance data that exemplifies the potential benefits of noncoherence, in terms of raw memory performance, as well as application speed.

Communicable Memory and Lazy Barriers for Bulk Synchronous Parallelism in BSPk

Communication and synchronization stand as the dual bottlenecks in the performance of parallel systems, and especially those that attempt to alleviate the programming burden by incurring overhead in these two domains. We formulate the notions of communicable memory and lazy barriers to help achieve efficient communication and synchronization. These concepts are developed in the context of BSPk, a toolkit library for programming networks of workstations|and other distributed memory architectures in general|based on the Bulk Synchronous Parallel (BSP) model. BSPk emphasizes efficiency in communication by minimizing local memory-to-memory copying, and in barrier synchronization by not forcing a process to wait unless it needs remote data. Both the message passing (MP) and distributed shared memory (DSM) programming styles are supported in BSPk. MP helps processes efficiently exchange short-lived unnamed data values, when the identity of either the sender or receiver is known to the other party. By contrast, DSM supports communication between processes that may be mutually anonymous, so long as they can agree on variable names in which to store shared temporary or long-lived data.

Informed Content Delivery Across Adaptive Overlay Networks

Overlay networks have emerged as a powerful and highly flexible method for delivering content. We study how to optimize throughput of large, multipoint transfers across richly connected overlay networks, focusing on the question of what to put in each transmitted packet. We first make the case for transmitting encoded content in this scenario, arguing for the digital fountain approach which enables end-hosts to efficiently restitute the original content of size n from a subset of any n symbols from a large universe of encoded symbols. Such an approach affords reliability and a substantial degree of application-level flexibility, as it seamlessly tolerates packet loss, connection migration, and parallel transfers. However, since the sets of symbols acquired by peers are likely to overlap substantially, care must be taken to enable them to collaborate effectively. We provide a collection of useful algorithmic tools for efficient estimation, summarization, and approximate reconciliation of sets of symbols between pairs of collaborating peers, all of which keep messaging complexity and computation to a minimum. Through simulations and experiments on a prototype implementation, we demonstrate the performance benefits of our informed content delivery mechanisms and how they complement existing overlay network architectures.

PreDA: Predicate Routing for DTN Architectures over MANET

We consider a Delay Tolerant Network (DTN) whose users (nodes) are connected by an underlying Mobile Ad hoc Network (MANET) substrate. Users can declaratively express high-level policy constraints on how “content” should be routed. For example, content can be directed through an intermediary DTN node for the purposes of preprocessing, authentication, etc., or content from a malicious MANET node can be dropped. To support such content routing at the DTN level, we implement Predicate Routing [1] where high-level constraints of DTN nodes are mapped into low-level routing predicates within the MANET nodes. Our testbed [2] uses a Linux system architecture with User Mode Linux [3] to emulate every DTN node with a DTN Reference Implementation code [4]. In our initial architecture prototype, we use the On Demand Distance Vector (AODV) routing protocol at the MANET level. We use the network simulator ns-2 (ns-emulation version) to simulate the wireless connectivity of both DTN and MANET nodes. Preliminary results show the efficient and correct operation of propagating routing predicates. For the application of content re-routing through an intermediary, as a side effect, results demonstrate the performance benefit of content re-routing that dynamically (on-demand) breaks the underlying end-to-end TCP connections into shorter-length TCP connections.

Distributed Packet Rewriting and its Application to Scalable Server Architectures

To construct high performance Web servers, system builders are increasingly turning to distributed designs. An important challenge that arises in distributed Web servers is the need to direct incoming connections to individual hosts. Previous methods for connection routing have employed a centralized node which handles all incoming requests. In contrast, we propose a distributed approach, called Distributed Packet Rewriting (DPR), in which all hosts of the distributed system participate in connection routing. We argue that this approach promises better scalability and fault-tolerance than the centralized approach. We describe our implementation of four variants of DPR and compare their performance. We show that DPR provides performance comparable to centralized alternatives, measured in terms of throughput and delay under the SPECweb96 benchmark. Finally, we argue that DPR is particularly attractive both for small scale systems and for systems following the emerging trend toward increasingly intelligent I/O subsystems.

From Parallel Sequence Representations to Calligraphic Control: A Conspiracy of Neural Circuits

Calligraphic writing presents a rich set of challenges to the human movement control system. These challenges include: initial learning, and recall from memory, of prescribed stroke sequences; critical timing of stroke onsets and durations; fine control of grip and contact forces; and letter-form invariance under voluntary size scaling, which entails fine control of stroke direction and amplitude during recruitment and derecruitment of musculoskeletal degrees of freedom. Experimental and computational studies in behavioral neuroscience have made rapid progress toward explaining the learning, planning and contTOl exercised in tasks that share features with calligraphic writing and drawing. This article summarizes computational neuroscience models and related neurobiological data that reveal critical operations spanning from parallel sequence representations to fine force control. Part one addresses stroke sequencing. It treats competitive queuing (CQ) models of sequence representation, performance, learning, and recall. Part two addresses letter size scaling and motor equivalence. It treats cursive handwriting models together with models in which sensory-motor tmnsformations are performed by circuits that learn inverse differential kinematic mappings. Part three addresses fine-grained control of timing and transient forces, by treating circuit models that learn to solve inverse dynamics problems.

Parallel Auditory Filtering By Sustained and Transient Channels Separates Coarticulated Vowels and Consonants

A neural model of peripheral auditory processing is described and used to separate features of coarticulated vowels and consonants. After preprocessing of speech via a filterbank, the model splits into two parallel channels, a sustained channel and a transient channel. The sustained channel is sensitive to relatively stable parts of the speech waveform, notably synchronous properties of the vocalic portion of the stimulus it extends the dynamic range of eighth nerve filters using coincidence deteectors that combine operations of raising to a power, rectification, delay, multiplication, time averaging, and preemphasis. The transient channel is sensitive to critical features at the onsets and offsets of speech segments. It is built up from fast excitatory neurons that are modulated by slow inhibitory interneurons. These units are combined over high frequency and low frequency ranges using operations of rectification, normalization, multiplicative gating, and opponent processing. Detectors sensitive to frication and to onset or offset of stop consonants and vowels are described. Model properties are characterized by mathematical analysis and computer simulations. Neural analogs of model cells in the cochlear nucleus and inferior colliculus are noted, as are psychophysical data about perception of CV syllables that may be explained by the sustained transient channel hypothesis. The proposed sustained and transient processing seems to be an auditory analog of the sustained and transient processing that is known to occur in vision.

Fast Learning VIEWNET Architectures for Recognizing 3-D Objects from Multiple 2-D Views

The recognition of 3-D objects from sequences of their 2-D views is modeled by a family of self-organizing neural architectures, called VIEWNET, that use View Information Encoded With NETworks. VIEWNET incorporates a preprocessor that generates a compressed but 2-D invariant representation of an image, a supervised incremental learning system that classifies the preprocessed representations into 2-D view categories whose outputs arc combined into 3-D invariant object categories, and a working memory that makes a 3-D object prediction by accumulating evidence from 3-D object category nodes as multiple 2-D views are experienced. The simplest VIEWNET achieves high recognition scores without the need to explicitly code the temporal order of 2-D views in working memory. Working memories are also discussed that save memory resources by implicitly coding temporal order in terms of the relative activity of 2-D view category nodes, rather than as explicit 2-D view transitions. Variants of the VIEWNET architecture may also be used for scene understanding by using a preprocessor and classifier that can determine both What objects are in a scene and Where they are located. The present VIEWNET preprocessor includes the CORT-X 2 filter, which discounts the illuminant, regularizes and completes figural boundaries, and suppresses image noise. This boundary segmentation is rendered invariant under 2-D translation, rotation, and dilation by use of a log-polar transform. The invariant spectra undergo Gaussian coarse coding to further reduce noise and 3-D foreshortening effects, and to increase generalization. These compressed codes are input into the classifier, a supervised learning system based on the fuzzy ARTMAP algorithm. Fuzzy ARTMAP learns 2-D view categories that are invariant under 2-D image translation, rotation, and dilation as well as 3-D image transformations that do not cause a predictive error. Evidence from sequence of 2-D view categories converges at 3-D object nodes that generate a response invariant under changes of 2-D view. These 3-D object nodes input to a working memory that accumulates evidence over time to improve object recognition. ln the simplest working memory, each occurrence (nonoccurrence) of a 2-D view category increases (decreases) the corresponding node's activity in working memory. The maximally active node is used to predict the 3-D object. Recognition is studied with noisy and clean image using slow and fast learning. Slow learning at the fuzzy ARTMAP map field is adapted to learn the conditional probability of the 3-D object given the selected 2-D view category. VIEWNET is demonstrated on an MIT Lincoln Laboratory database of l28x128 2-D views of aircraft with and without additive noise. A recognition rate of up to 90% is achieved with one 2-D view and of up to 98.5% correct with three 2-D views. The properties of 2-D view and 3-D object category nodes are compared with those of cells in monkey inferotemporal cortex.