Task Assignment in a Distributed System: Improving Performance by Unbalancing Load

We consider the problem of task assignment in a distributed system (such as a distributed Web server) in which task sizes are drawn from a heavy-tailed distribution. Many task assignment algorithms are based on the heuristic that balancing the load at the server hosts will result in optimal performance. We show this conventional wisdom is less true when the task size distribution is heavy-tailed (as is the case for Web file sizes). We introduce a new task assignment policy, called Size Interval Task Assignment with Variable Load (SITA-V). SITA-V purposely operates the server hosts at different loads, and directs smaller tasks to the lighter-loaded hosts. The result is that SITA-V provably decreases the mean task slowdown by significant factors (up to 1000 or more) where the more heavy-tailed the workload, the greater the improvement factor. We evaluate the tradeoff between improvement in slowdown and increase in waiting time in a system using SITA-V, and show conditions under which SITA-V represents a particularly appealing policy. We conclude with a discussion of the use of SITA-V in a distributed Web server, and show that it is attractive because it has a simple implementation which requires no communication from the server hosts back to the task router.

Program Representation Size in an Intermediate Language with Intersection and Union Types

The CIL compiler for core Standard ML compiles whole programs using a novel typed intermediate language (TIL) with intersection and union types and flow labels on both terms and types. The CIL term representation duplicates portions of the program where intersection types are introduced and union types are eliminated. This duplication makes it easier to represent type information and to introduce customized data representations. However, duplication incurs compile-time space costs that are potentially much greater than are incurred in TILs employing type-level abstraction or quantification. In this paper, we present empirical data on the compile-time space costs of using CIL as an intermediate language. The data shows that these costs can be made tractable by using sufficiently fine-grained flow analyses together with standard hash-consing techniques. The data also suggests that non-duplicating formulations of intersection (and union) types would not achieve significantly better space complexity.

A Spectrum of TCP-Friendly Window-Based Congestion Control Algorithms

The increased diversity of Internet application requirements has spurred recent interests in transport protocols with flexible transmission controls. In window-based congestion control schemes, increase rules determine how to probe available bandwidth, whereas decrease rules determine how to back off when losses due to congestion are detected. The parameterization of these control rules is done so as to ensure that the resulting protocol is TCP-friendly in terms of the relationship between throughput and loss rate. In this paper, we define a new spectrum of window-based congestion control algorithms that are TCP-friendly as well as TCP-compatible under RED. Contrary to previous memory-less controls, our algorithms utilize history information in their control rules. Our proposed algorithms have two salient features: (1) They enable a wider region of TCP-friendliness, and thus more flexibility in trading off among smoothness, aggressiveness, and responsiveness; and (2) they ensure a faster convergence to fairness under a wide range of system conditions. We demonstrate analytically and through extensive ns simulations the steady-state and transient behaviors of several instances of this new spectrum of algorithms. In particular, SIMD is one instance in which the congestion window is increased super-linearly with time since the detection of the last loss. Compared to recently proposed TCP-friendly AIMD and binomial algorithms, we demonstrate the superiority of SIMD in: (1) adapting to sudden increases in available bandwidth, while maintaining competitive smoothness and responsiveness; and (2) rapidly converging to fairness and efficiency.

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.

Safe Composition of Web Communication Protocols for Extensible Edge Services

As new multi-party edge services are deployed on the Internet, application-layer protocols with complex communication models and event dependencies are increasingly being specified and adopted. To ensure that such protocols (and compositions thereof with existing protocols) do not result in undesirable behaviors (e.g., livelocks) there needs to be a methodology for the automated checking of the "safety" of these protocols. In this paper, we present ingredients of such a methodology. Specifically, we show how SPIN, a tool from the formal systems verification community, can be used to quickly identify problematic behaviors of application-layer protocols with non-trivial communication models—such as HTTP with the addition of the "100 Continue" mechanism. As a case study, we examine several versions of the specification for the Continue mechanism; our experiments mechanically uncovered multi-version interoperability problems, including some which motivated revisions of HTTP/1.1 and some which persist even with the current version of the protocol. One such problem resembles a classic degradation-of-service attack, but can arise between well-meaning peers. We also discuss how the methods we employ can be used to make explicit the requirements for hardening a protocol's implementation against potentially malicious peers, and for verifying an implementation's interoperability with the full range of allowable peer behaviors.

Fast Approximate Reconciliation of Set Differences

We present new, simple, efficient data structures for approximate reconciliation of set differences, a useful standalone primitive for peer-to-peer networks and a natural subroutine in methods for exact reconciliation. In the approximate reconciliation problem, peers A and B respectively have subsets of elements SA and SB of a large universe U. Peer A wishes to send a short message M to peer B with the goal that B should use M to determine as many elements in the set SB–SA as possible. To avoid the expense of round trip communication times, we focus on the situation where a single message M is sent. We motivate the performance tradeoffs between message size, accuracy and computation time for this problem with a straightforward approach using Bloom filters. We then introduce approximation reconciliation trees, a more computationally efficient solution that combines techniques from Patricia tries, Merkle trees, and Bloom filters. We present an analysis of approximation reconciliation trees and provide experimental results comparing the various methods proposed for approximate reconciliation.

Validating Arbitrarily Large Network Protocol Compositions with Finite Computation

Formal tools like finite-state model checkers have proven useful in verifying the correctness of systems of bounded size and for hardening single system components against arbitrary inputs. However, conventional applications of these techniques are not well suited to characterizing emergent behaviors of large compositions of processes. In this paper, we present a methodology by which arbitrarily large compositions of components can, if sufficient conditions are proven concerning properties of small compositions, be modeled and completely verified by performing formal verifications upon only a finite set of compositions. The sufficient conditions take the form of reductions, which are claims that particular sequences of components will be causally indistinguishable from other shorter sequences of components. We show how this methodology can be applied to a variety of network protocol applications, including two features of the HTTP protocol, a simple active networking applet, and a proposed web cache consistency algorithm. We also doing discuss its applicability to framing protocol design goals and to representing systems which employ non-model-checking verification methodologies. Finally, we briefly discuss how we hope to broaden this methodology to more general topological compositions of network applications.

Using Markets and Spam to Combat Malware

We propose an economic mechanism to reduce the incidence of malware that delivers spam. Earlier research proposed attention markets as a solution for unwanted messages, and showed they could provide more net benefit than alternatives such as filtering and taxes. Because it uses a currency system, Attention Bonds faces a challenge. Zombies, botnets, and various forms of malware might steal valuable currency instead of stealing unused CPU cycles. We resolve this problem by taking advantage of the fact that the spam-bot problem has been reduced to financial fraud. As such, the large body of existing work in that realm can be brought to bear. By drawing an analogy between sending and spending, we show how a market mechanism can detect and prevent spam malware. We prove that by using a currency (i) each instance of spam increases the probability of detecting infections, and (ii) the value of eradicating infections can justify insuring users against fraud. This approach attacks spam at the source, a virtue missing from filters that attack spam at the destination. Additionally, the exchange of currency provides signals of interest that can improve the targeting of ads. ISPs benefit from data management services and consumers benefit from the higher average value of messages they receive. We explore these and other secondary effects of attention markets, and find them to offer, on the whole, attractive economic benefits for all – including consumers, advertisers, and the ISPs.

Connection Scheduling in Web Servers

Under high loads, a Web server may be servicing many hundreds of connections concurrently. In traditional Web servers, the question of the order in which concurrent connections are serviced has been left to the operating system. In this paper we ask whether servers might provide better service by using non-traditional service ordering. In particular, for the case when a Web server is serving static files, we examine the costs and benefits of a policy that gives preferential service to short connections. We start by assessing the scheduling behavior of a commonly used server (Apache running on Linux) with respect to connection size and show that it does not appear to provide preferential service to short connections. We then examine the potential performance improvements of a policy that does favor short connections (shortest-connection-first). We show that mean response time can be improved by factors of four or five under shortest-connection-first, as compared to an (Apache-like) size-independent policy. Finally we assess the costs of shortest-connection-first scheduling in terms of unfairness (i.e., the degree to which long connections suffer). We show that under shortest-connection-first scheduling, long connections pay very little penalty. This surprising result can be understood as a consequence of heavy-tailed Web server workloads, in which most connections are small, but most server load is due to the few large connections. We support this explanation using analysis.

Distributed Selfish Coaching

Although cooperation generally increases the amount of resources available to a community of nodes, thus improving individual and collective performance, it also allows for the appearance of potential mistreatment problems through the exposition of one node's resources to others. We study such concerns by considering a group of independent, rational, self-aware nodes that cooperate using on-line caching algorithms, where the exposed resource is the storage at each node. Motivated by content networking applications -- including web caching, CDNs, and P2P -- this paper extends our previous work on the on-line version of the problem, which was conducted under a game-theoretic framework, and limited to object replication. We identify and investigate two causes of mistreatment: (1) cache state interactions (due to the cooperative servicing of requests) and (2) the adoption of a common scheme for cache management policies. Using analytic models, numerical solutions of these models, as well as simulation experiments, we show that on-line cooperation schemes using caching are fairly robust to mistreatment caused by state interactions. To appear in a substantial manner, the interaction through the exchange of miss-streams has to be very intense, making it feasible for the mistreated nodes to detect and react to exploitation. This robustness ceases to exist when nodes fetch and store objects in response to remote requests, i.e., when they operate as Level-2 caches (or proxies) for other nodes. Regarding mistreatment due to a common scheme, we show that this can easily take place when the "outlier" characteristics of some of the nodes get overlooked. This finding underscores the importance of allowing cooperative caching nodes the flexibility of choosing from a diverse set of schemes to fit the peculiarities of individual nodes. To that end, we outline an emulation-based framework for the development of mistreatment-resilient distributed selfish caching schemes. Our framework utilizes a simple control-theoretic approach to dynamically parameterize the cache management scheme. We show performance evaluation results that quantify the benefits from instantiating such a framework, which could be substantial under skewed demand profiles.

Distributed Placement of Service Facilities in Large-Scale Networks

The effectiveness of service provisioning in largescale networks is highly dependent on the number and location of service facilities deployed at various hosts. The classical, centralized approach to determining the latter would amount to formulating and solving the uncapacitated k-median (UKM) problem (if the requested number of facilities is fixed), or the uncapacitated facility location (UFL) problem (if the number of facilities is also to be optimized). Clearly, such centralized approaches require knowledge of global topological and demand information, and thus do not scale and are not practical for large networks. The key question posed and answered in this paper is the following: "How can we determine in a distributed and scalable manner the number and location of service facilities?" We propose an innovative approach in which topology and demand information is limited to neighborhoods, or balls of small radius around selected facilities, whereas demand information is captured implicitly for the remaining (remote) clients outside these neighborhoods, by mapping them to clients on the edge of the neighborhood; the ball radius regulates the trade-off between scalability and performance. We develop a scalable, distributed approach that answers our key question through an iterative reoptimization of the location and the number of facilities within such balls. We show that even for small values of the radius (1 or 2), our distributed approach achieves performance under various synthetic and real Internet topologies that is comparable to that of optimal, centralized approaches requiring full topology and demand information.

Implications of Selfish Neighbor Selection in Overlay Networks

In a typical overlay network for routing or content sharing, each node must select a fixed number of immediate overlay neighbors for routing traffic or content queries. A selfish node entering such a network would select neighbors so as to minimize the weighted sum of expected access costs to all its destinations. Previous work on selfish neighbor selection has built intuition with simple models where edges are undirected, access costs are modeled by hop-counts, and nodes have potentially unbounded degrees. However, in practice, important constraints not captured by these models lead to richer games with substantively and fundamentally different outcomes. Our work models neighbor selection as a game involving directed links, constraints on the number of allowed neighbors, and costs reflecting both network latency and node preference. We express a node's "best response" wiring strategy as a k-median problem on asymmetric distance, and use this formulation to obtain pure Nash equilibria. We experimentally examine the properties of such stable wirings on synthetic topologies, as well as on real topologies and maps constructed from PlanetLab and AS-level Internet measurements. Our results indicate that selfish nodes can reap substantial performance benefits when connecting to overlay networks composed of non-selfish nodes. On the other hand, in overlays that are dominated by selfish nodes, the resulting stable wirings are optimized to such great extent that even non-selfish newcomers can extract near-optimal performance through naive wiring strategies.

Biologically Inspired Approaches to Automated Feature Extraction and Target Recognition

Ongoing research at Boston University has produced computational models of biological vision and learning that embody a growing corpus of scientific data and predictions. Vision models perform long-range grouping and figure/ground segmentation, and memory models create attentionally controlled recognition codes that intrinsically cornbine botton-up activation and top-down learned expectations. These two streams of research form the foundation of novel dynamically integrated systems for image understanding. Simulations using multispectral images illustrate road completion across occlusions in a cluttered scene and information fusion from incorrect labels that are simultaneously inconsistent and correct. The CNS Vision and Technology Labs (cns.bu.edulvisionlab and cns.bu.edu/techlab) are further integrating science and technology through analysis, testing, and development of cognitive and neural models for large-scale applications, complemented by software specification and code distribution.

Fusion Artmap: A Neural Network Architecture for Multi-Channel Data Fusion and Classification

Fusion ARTMAP is a self-organizing neural network architecture for multi-channel, or multi-sensor, data fusion. Single-channel Fusion ARTMAP is functionally equivalent to Fuzzy ART during unsupervised learning and to Fuzzy ARTMAP during supervised learning. The network has a symmetric organization such that each channel can be dynamically configured to serve as either a data input or a teaching input to the system. An ART module forms a compressed recognition code within each channel. These codes, in turn, become inputs to a single ART system that organizes the global recognition code. When a predictive error occurs, a process called paraellel match tracking simultaneously raises vigilances in multiple ART modules until reset is triggered in one of them. Parallel match tracking hereby resets only that portion of the recognition code with the poorest match, or minimum predictive confidence. This internally controlled selective reset process is a type of credit assignment that creates a parsimoniously connected learned network. Fusion ARTMAP's multi-channel coding is illustrated by simulations of the Quadruped Mammal database.

The What-And-Where Filter: A Spatial Mapping Neural Network for Object Recognition and Image Understanding

The What-and-Where filter forms part of a neural network architecture for spatial mapping, object recognition, and image understanding. The Where fllter responds to an image figure that has been separated from its background. It generates a spatial map whose cell activations simultaneously represent the position, orientation, ancl size of all tbe figures in a scene (where they are). This spatial map may he used to direct spatially localized attention to these image features. A multiscale array of oriented detectors, followed by competitve and interpolative interactions between position, orientation, and size scales, is used to define the Where filter. This analysis discloses several issues that need to be dealt with by a spatial mapping system that is based upon oriented filters, such as the role of cliff filters with and without normalization, the double peak problem of maximum orientation across size scale, and the different self-similar interpolation properties across orientation than across size scale. Several computationally efficient Where filters are proposed. The Where filter rnay be used for parallel transformation of multiple image figures into invariant representations that are insensitive to the figures' original position, orientation, and size. These invariant figural representations form part of a system devoted to attentive object learning and recognition (what it is). Unlike some alternative models where serial search for a target occurs, a What and Where representation can he used to rapidly search in parallel for a desired target in a scene. Such a representation can also be used to learn multidimensional representations of objects and their spatial relationships for purposes of image understanding. The What-and-Where filter is inspired by neurobiological data showing that a Where processing stream in the cerebral cortex is used for attentive spatial localization and orientation, whereas a What processing stream is used for attentive object learning and recognition.