Search

A generic objective for a sensor network application is thegathering of data from a field of sensors. Because energy is often scarce insensor networks, many techniques have been proposed to reduce data size withinthe network. These techniques either ignore the accuracy of the resulting data,or more often, provide no means for applications to control the resultingaccuracy. However in many cases, applications have a quantitative requirementfor sensor data accuracy, and the underlying system should meet thatefficiently. In this paper, we describe a distributed algorithm thatapproximates and gathers data in an energy-efficient manner and strictlysatisfies an application-provided accuracy requirement. This approximation isbased on a hybrid data representation based on linear regression. Adistinguishing feature of the proposed algorithm is that it absolutely does notrequire any models on statistical properties of data and noise, and needs onlyfew general assumptions on sensor node topology. This feature enables thealgorithm to serve as a general-purpose mechanism that can be widely used inmany scenarios for data gathering-type applications. Simulation experimentswith data traces from real environmental data show that it leverages theaccuracy requirement to significantly reduce energy consumption.Pre-2018 CSE ID: CS2006-0854

Emerging large-scale scientific applications involve massive,distributed, shared data collections (petabytes), and require robust, highperformance for read-dominated workloads. Achieving robust performance (lowvariability) in storage systems is difficult. We propose RobuSTore, a novelstorage technique, which combines erasure codes and speculative access toreduce performance variability and increase performance. RobuSTore useserasure codes to add flexible redundancy then spreads the encoded data across alarge number of disks. Speculative access to the redundant data from multipledisks enables application requests to be satisfied with only early-arrivingblocks, reducing performance dependence on the behavior of individual disks.We present the design and an evaluation of RobuSTore which shows improvedrobustness, reducing the standard deviation of access latencies by as much as5-fold vs. traditional RAID. In addition, RobuSTore improves access bandwidthby as much as 15-fold. A typical 1 GB read from 64 disks has average latency of2 seconds with standard deviation of only 0.5 seconds or 25%. RobuSToresecures these benefits at the cost of a 2-3x storage capacity overhead and~1.5x network and disk I/O overhead.Pre-2018 CSE ID: CS2006-0851

Media companies and other organizations with large amounts ofdigital content require transfer of extremely large files in a short time froma single source to a collection of geographically dispersed destinations. Dueto the high cost of terrestrial networks of sufficient scope and bandwidth,satellite networks are the most common means for performing such transfers.However, satellite file transfer relies on expensive error correction based ona combination of forward error correction and whole-file retransmission.This paper presents a hybrid solution that combines the advantages of satelliteand terrestrial networks to provide cost-effective reliable file transfer.Specifically, we propose a new peer-to-peer (P2P) scheme that exploits fastterrestrial networks and the availability of multiple receivers to recover fromhigh loss rates (5% or more) in near real-time (latency

A critical technology for Grid computing is the ability to describe,select, and bind appropriate resources for synchronous use by applications. OurVirtual Grid Description Language (vgDL) allows applications to describe andmanage resources conveniently via application-level abstractions and enables anovel approach to application-driven resource management called "finding andbinding". We explore the viability of this strategy to enables applications toobtain complex resource collections in competitive resource environments. Ourevaluation shows that combined resource "selection and binding" can scale wellto millions of hosts, identifying good matches in a few seconds for the mostcomplex resource requests. The combined selection and binding improves successrates for complex requests and the advantage increases for more complexrequests and higher resource competition. Combined Selection and Binding candouble the resource utilization (from 30% to 70%) at which synchronous resourceallocation and use across as many as sixteen resource managers is possible.Pre-2018 CSE ID: CS2005-0825

Simple resource specification, resource selection, and effectivebinding are critical capabilities for Grid middleware. We describe the VirtualGrid, an abstraction for dynamic grid applications to deal with complexresource environments. Elements of the Virtual Grid include a novel resourcedescription language (vgDL) and a resource selection and binding component(vgFAB), which accepts a vgDL specification and returns a Virtual Grid, thatis, a set of selected and bound resources. The goals of vgFAB are efficiency,scalability, robustness to high resource contention, and the ability to produceresults with quantifiable high quality. We present the design of vgDL, showinghow it captures application-level resource abstractions using resourceaggregates and connectivity amongst them.Pre-2018 CSE ID: CS2005-0817

Simple resource specification, resource selection, and effectivebinding are critical capabilities for Grid middleware. We describe the VirtualGrid, an abstraction for dynamic grid applications to deal with complexresource environments. Elements of the Virtual Grid include a novel resourcedescription language (vgDL) and a resource selection and binding component(vgFAB), which accepts a vgDL specification and returns a Virtual Grid, thatis, a set of selected and bound resources. The goals of vgFAB are efficiency,scalability, robustness to high resource contention, and the ability to produceresults with quantifiable high quality. We present the design of vgDL, showinghow it captures application-level resource abstractions using resourceaggregates and connectivity amongst them. We present and evaluate a prototypeimplementation of vgFAB. Our results show that resource selection and bindingfor virtual grids of 10,000's of resources can scale up to grids with millionsof resources, identifying good matches in less than one second. Further, thesematches have quantifiable quality, enabling applications to have highconfidence in the results. We demonstrate the effectiveness of our combinedselection and binding approach in the presence of resource contention, showingthe robust selection and binding can be achieved at moderate cost.Pre-2018 CSE ID: CS2004-0809

We examine the properties and performance of LT Codes for use in ahigh performance distributed file system. Different variations to thealgorithm were created to achieve key characteristics, such as speed and datareliability. Finally, we evaluate parameter choices in order to balance thesystem and network loads.Pre-2018 CSE ID: CS2004-0798

Overlay networks (proxy networks) have been used as a communicationinfrastructure to allow applications to communicate with users withoutrevealing their IP addresses. Such proxy networks are used to enhanceapplication security; including protecting applications from direct attacks andinfrastructure Denial-of-Service attacks. However, the conditions under whichsuch approaches can hide application location are not well understood. To shedlight on this question, we develop a formal framework for proxy networkapproach to location-hiding which encompasses most of the proposed approaches.This framework is used to analyze the effectiveness of location-hiding:characterizing how attacks, defenses, and correlated host vulnerabilitiesaffect feasibility. We find that existing approaches employing staticstructure (e.g. SOS and I3) cannot hide application location because attackersgain information monotonically and in a short period of time penetrate theproxy network. However, we find that adding defenses such as networkreconfiguration or migration, which invalidate the information attackers have,makes location-hiding feasible for resisting penetration attacks. We alsocharacterize when proxy networks are effective and stable in location-hiding.In these systems, proxy network depth and reconfiguration rates are criticalfactors for effectiveness. These results provide both deeper understanding ofthe location-hiding problem and guidelines for proxy network design.Pre-2018 CSE ID: CS2004-0788

We describe a shared perspective and assumptions looking up and downthe OptIPuter system software stack. The discussion clearly defines andjustifies the assumptions and model for each perspective, starting with thenetwork hardware architecture, the integrating model of distributed virtualcomputers and the supporting technologies of group transport protocol,multi-endpoint communication, and shared lambda-grid storage.Pre-2018 CSE ID: CS2004-0786

Desktop distributed computing allows companies to exploit the idlecycles on pervasive desktop PC systems to increase the available computingpower by orders of magnitude (10x - 1000x). Applications are submitted,distributed, and run on a grid of desktop PCs. Since the applications may bemalformed, or malicious, the key challenges for a desktop grid are how to 1)control the distributed computing application's resource usage and behavior asit runs on the desktop PC, 2) provide protection for the distributedapplication's program and its data, and 3) provide a virtualized executionenvironment, decoupling correct application execution from desktopconfiguration heterogeneity. We describe the Entropia Virtual Machine, and thesolutions it embodies for each of these challenges. We first outline thedistinct requirements for a virtual machine in a desktop grid environment,which differ from web-based and hardware-level virtual machines. We thendescribe how the Entropia Virtual Machine (EVM) uses binary rewriting to solvethese problems, guaranteeing the usability and protection of the desktopmachine in the face of malicious or malformed application code. We also showthat our approach provides security for the applications and their data. Wedetail how the virtual machine supports a virtualized execution environment.We present performance results for the EVM which show its performance overheadis modest, less than 6% for the applications studied.Pre-2018 CSE ID: CS2003-0773

Accurate and continuous monitoring and profiling are importantissues of performance tuning and scheduling optimization. In desktop gridsystems based on sandboxing techniques these issues are particularlychallenging because (1) subjobs inside sandboxes are executed in a virtualenvironment and (2) sandboxes are usually reset to an initial (empty) state ateach subjob termination. To address this problem, we present in this paperDGMonitor, a monitoring tool to build a global, accurate and continuous view ofreal resource utilization for desktop grids based on sandboxing techniques. Ourmonitoring tool provides unobtrusive and reliable performance measures, uses asimple performance data model and is easy to use. Our work demonstrates thatDGMonitor can easily take over the monitoring of large desktop grids (up to12000 workers) maintaining low load in terms of resource consumption due to themonitoring process (less than 0.1%) on desktop PCs. Although we use DGMonitorwith the Entropia DGrid platform, it can be easily integrated in other desktopgrids and its data can be used as an information source for existinginformation services for performance tuning and scheduling optimization.Keywords: Performance monitoring and profiling, desktop grids, sandboxingtechniques, distributed computing.Pre-2018 CSE ID: CS2003-0770

Desktop resources are attractive for running compute-intensivedistributed applications. Several systems that aggregate these resources indesktop grids have been developed. While these systems have been successfullyused for a wide variety of high throughput applications there has been littleinsight into the detailed temporal structure of CPU availability of desktopgrid resources. Yet, this structure is critical to characterize the utility ofdesktop grid platforms for both task parallel and even data parallelapplications. We address the following questions: (i) What are the temporalcharacteristics of desktop CPU availability in an enterprise setting? (ii) Howdo these characteristics affect the utility of desktop grids? (iii) Based onthese characteristics, can we construct a model of server "equivalents" forthe desktop grids, which can be used to predict application performance? Wepresent measurements of an enterprise desktop grid with over 220 hosts runningthe Entropia commercial desktop grid software. We utilize these measurementsto characterize CPU availability and develop a performance model for desktopgrid applications for various task granularities, showing that there is anoptimal task size. We then introduce a new metric, cluster equivalence, whichwe use to quantify the utility of the desktop grid relative to that of adedicated cluster.Pre-2018 CSE ID: CS2003-0769

Abstract Treponema pallidum subsp. pallidum, the causative agent of syphilis, is a highly invasive spirochete pathogen that uses the vasculature to disseminate throughout the body. Identification of bacterial factors promoting dissemination is crucial for syphilis vaccine development. An important step in dissemination is bacterial adhesion to blood vessel surfaces, a process mediated by bacterial proteins that can withstand forces imposed on adhesive bonds by blood flow (vascular adhesins). The study of T. pallidum vascular adhesins is hindered by the uncultivable nature of this pathogen. We overcame these limitations by expressing T. pallidum adhesin Tp0751 (pallilysin) in an adhesion-attenuated strain of the cultivable spirochete Borrelia burgdorferi. Under fluid shear stress representative of conditions in postcapillary venules, Tp0751 restored bacterial-vascular interactions to levels similar to those observed for infectious B. burgdorferi and a gain-of-function strain expressing B. burgdorferi vascular adhesin BBK32. The strength and stability of Tp0751- and BBK32-dependent endothelial interactions under physiological shear stress were similar, although the mechanisms stabilizing these interactions were distinct. Tp0751 expression also permitted bacteria to interact with postcapillary venules in live mice as effectively as BBK32-expressing strains. These results demonstrate that Tp0751 can function as a vascular adhesin.

We propose a machine architecture that integrates programmable logic into key components of the system with the goal of customizing architectural mechanisms and policies to match an application. This approach presents an improvement over traditional approach of exploiting programmable logic as a separate co-processor by preserving machine usability through software and over traditional computer architecture by providing application-specific hardware assists. We present two case studies of architectural customization to enhance latency tolerance and efficiently utilize network bisection on multiprocessors for sparse matrix computations. We demonstrate that using application-specific hardware assists and policies can provide substantial improvements in performance on a per application basis. Based on these preliminary results, we propose that an application-driven machine customization provides a promising approach to achieve high performance and combat performance fragility.

The Scalable Systems Laboratory (SSL), part of the IRIS-HEP Software Institute, provides Institute participants and HEP software developers generally with a means to transition their R&D from conceptual toys to testbeds to production-scale prototypes. The SSL enables tooling, infrastructure, and services supporting innovation of novel analysis and data architectures, development of software elements and tool-chains, reproducible functional and scalability testing of service components, and foundational systems R&D for accelerated services developed by the Institute. The SSL is constructed with a core team having expertise in scale testing and deployment of services across a wide range of cyberinfrastructure. The core team embeds and partners with other areas in the Institute, and with LHC and other HEP development and operations teams as appropriate, to define investigations and required service deployment patterns. We describe the approach and experiences with early application deployments, including analysis platforms and intelligent data delivery systems.

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.comnet.2005.07.022 Byline: Ju Wang, Andrew A. Chien Keywords: Overlay network; Proxy network; Location-hiding; Security; Penetration attack; Correlated vulnerabilities Abstract: Overlay networks (proxy networks) have been used as a communication infrastructure to allow applications to communicate with users without revealing their IP addresses. Such proxy networks are used to enhance application security; including protecting applications from direct attacks and infrastructure Denial-of-Service attacks. However, the conditions under which such approaches can hide application location are not well understood. To shed light on this question, we develop a formal framework for the proxy network approach to location-hiding which encompasses most of the proposed approaches. It is used to characterize how attacks, defenses, and correlated host vulnerabilities affect the feasibility of location-hiding. We find that existing approaches employing static structures (e.g., SOS and I3) cannot hide application location because attackers gain information monotonically and quickly penetrate the proxy network. However, adding defenses, such as proxy network reconfiguration or migration, which invalidate the information attackers have, makes location-hiding feasible against penetration attacks. Proxy-network depth and reconfiguration rates are critical factors for effectiveness. Furthermore, correlated vulnerabilities in many cases jeopardize location-hiding; however, by exploiting host diversity and intelligent proxy-network construction, the negative impact of correlation can be mitigated and location-hiding can be achieved. These results provide deeper understanding of the location-hiding problem and guidelines for proxy-network design. Author Affiliation: Department of Computer Science and Engineering, University of California -- San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0114, United States

A simple synchronization mechanism called Feedback based Synchronization (FBS), which uses built- in flow control signals was proposed. Results reveal that the maximum bounded skew of FBS is 5.79 (mu)sec or less over all the experiments and hence it is concluded that FBS is a feasible synchronization in system area networks.

A considerable amount of research and engineering went into designing proxy applications, which represent common high-performance computing (HPC) workloads, to co-design and evaluate the current generation of supercomputers, e.g., RIKEN's supercomputer Fugaku, ANL's Aurora, or ORNL's Frontier. This process was necessary to standardize the procurement while avoiding duplicated effort at each HPC center to develop their own benchmarks. Unfortunately, proxy applications force HPC centers and providers (vendors) into an undesirable state of rigidity, in contrast to the fast-moving trends of current technology and future heterogeneity. To accommodate an extremely heterogeneous future, we have to reconsider how to co-design supercomputers during the next decade, and avoid repeating past mistakes. [ABSTRACT FROM AUTHOR]

Routing networks are important to the performance of parallel machines. Adaptive routers have usually been assessed on the basis of achieved network throughput. A parameterized cost model for router performance is presented, distinguished by router delay and flow control time. From design studies it is clear that absolute addressing should be reconsidered, and the complexity of adaptive routes increases setup delay and flow control cycle times.

Compressionless Routing (CR) is a new adaptive routing framework that supports both adaptive and fault-tolerant routing for various network topologies. CR uses the fine-grained flow control and backpressure of wormhole routing to communicate routing status and error conditions to network interfaces. Among its advantages are simpler router designs, deadlock-free adaptive routing in torus networks with no virtual channels, order-preserving message transmission and permanent faults tolerance.