Papers by Joseph Pasquale
Abstract: Internet computing is highly network-integrated computing on existing networks of compu... more Abstract: Internet computing is highly network-integrated computing on existing networks of computers, done on the grand scale of the Internet. Supporting Internet computing will require advances in network communications architectures. I outline three research directions that I believe are critical: resource sharing based on market-economic principles; communication of agents (ie where the basic type of message is a program and not simply data); and provisions for ranges of qualities of service.
Abstract Expert systems can be used effectively to manage distributed computer systems which are ... more Abstract Expert systems can be used effectively to manage distributed computer systems which are based on decentralized control of shared resources. These distributed systems can exhibit high reliability and performance. Yet, designing such systems pose formidable problems. These problems involve real-time distributed decision-making where decision-makers do not know with full certainty the state of remote nodes.
Abstract A model is presented where uncertainty exists concerning whether agents work together in... more Abstract A model is presented where uncertainty exists concerning whether agents work together in a coalition or work alone, capturing the uncertainty for group behavior in political, economic or distributed computing systems. The uncertainty is modeled as a probability of group formation and the performance of the system is modeled as payoffs in a simple game matrix. Both the level of uncertainty and the game payoffs determine the type of equilibria in the actions of the decision-makers.
Abstract: We investigated new system software architectures that support mobile code, based on ne... more Abstract: We investigated new system software architectures that support mobile code, based on network (and distributed) computing. The primary goal was to support an application's reliance on network resources, rather than local ones (which are often scarce in defense mobile-computing situations), to meet their communication and computational demands. We have developed a powerful extension of the client/server model, which we call the Extended Client/Server (ECS) model of distributed computing.
Abstract To support application-specific quality of service for hosted services, a client of a co... more Abstract To support application-specific quality of service for hosted services, a client of a compute utility requires the ability to schedule the processor resources supplied to its service. We present a user-level scheduling framework that operates in tandem with a standard kernel scheduler to support user-level policies for sharing processor resources. The scheduler operates by sampling the resource consumption of processes and limiting which processes are eligible for scheduling by the kernel.
Abstract New I/O devices with data rates ranging from 10 to 100 Mbytes per second are becoming av... more Abstract New I/O devices with data rates ranging from 10 to 100 Mbytes per second are becoming available for personal computers and workstations. Along with continual improvements in processor, memory, and bus technology, these devices have enabled I/O-intensive applications for desktop computing that require input, processing, and output of very large amounts of data.
Abstract We investigate how delayed communication affects the dynamic formation of groups in dist... more Abstract We investigate how delayed communication affects the dynamic formation of groups in distributed systems, where all decision-making agents join the same group because each expects to improve its own performance. For example, distributed job schedulers may form a group to utilize the idle resources of other members within the group. Forming a group is a search problem and we examine agents that use the feedback mechanism of stochastic learning automata to carry out this search.
Abstract The concept of agents--programs that are capable of transporting themselves across a [he... more Abstract The concept of agents--programs that are capable of transporting themselves across a [heterogeneous] network to execute and return results--is a fascinating if troubled area of research. While theoretical advantages of agents have been well-established, few agent-based applications have been commercially successful. We argue that the lack of applications stems from a lack of understanding essential agent usage patterns.
In this report we discuss the strategic directions and challenges in the management and use of st... more In this report we discuss the strategic directions and challenges in the management and use of storage systems| those components of computer systems responsible for the storage and retrieval of data.
Abstract In this paper, we investigate the roles of replication vs. repair to achieve durability ... more Abstract In this paper, we investigate the roles of replication vs. repair to achieve durability in large-scale distributed storage systems. Specifically, we address the fundamental questions: How does the lifetime of an object depend on the degree of replication and rate of repair, and how is lifetime maximized when there is a constraint on resources?
ABSTRACT Source-based dithering is a set of techniques designed to maximize the performance of re... more ABSTRACT Source-based dithering is a set of techniques designed to maximize the performance of real-time networked digital video systems that encode and decode video entirely in software. Usually frame grabber hardware presents frames in a 24 bit per pixel (bpp) format. However, most hosts are only equipped with single or eight bit deep displays and thus the color depth of the video must be reduced at some point.
Abstract Virtual Network Computing, or VNC, is a popular thin client application used to access f... more Abstract Virtual Network Computing, or VNC, is a popular thin client application used to access files and applications on remote computers. However, VNC can suffer from significant losses in throughput when there is high latency between the client and server. These losses become especially apparent in the case of video, where updates are both large and frequent. In this work, we present a Message Accelerator proxy for VNC.
Abstract Expert systems which maintain knowledge about objects whose attributes are time-variant ... more Abstract Expert systems which maintain knowledge about objects whose attributes are time-variant must have an awareness of time. This awareness can be made manifest by incorporating time in the quantification of uncertainty of aging knowledge about such objects. Many expert systems use some method to quantify the degree of belief, or uncertainty, of their knowledge.
Abstract Networking software is a growing bottleneck in modern workstations, particularly for hig... more Abstract Networking software is a growing bottleneck in modern workstations, particularly for high throughput applications such as networked digital video. We measure various components of the UDP/IP protocol stack in a DECstation 5000/200 running Ultrix 4.2 a, and quantify the way in which checksumming and copying dominate the processing time for high throughput applications.
We describe the concept of the relocatable continuous media filter. The novelty of these filters ... more We describe the concept of the relocatable continuous media filter. The novelty of these filters is how they can propagate over a dissemination tree in a network. We describe the filter propagation protocol to achieve this. Execution of filters inside a network allows the network to be viewed in a novel way, as a processor with its instruction set being the various types of available filters.
We report performance measurements of Internet connections between five Sequoia 2000 sites. Throu... more We report performance measurements of Internet connections between five Sequoia 2000 sites. Throughput and delay statistics are presented for various message sizes and for both daytime and nighttime. The highest throughput observed was 85 KB/s between UCSD and UCLA at night and the lowest was 1 KB/s between UCSD and DWR during the day.
Abstract At the Computer Systems Laboratory at UCSD, we conduct research in operating systems, co... more Abstract At the Computer Systems Laboratory at UCSD, we conduct research in operating systems, communication networks, performance evaluation, and distributed systems. This document reviews the current research within the Computer Systems Laboratory. We describe our involvement in the Sequoia 2000 Project, and we describe our work in the areas listed above. The nal section contains descriptions of research projects undertaken by our graduate students.
Abstract: P2P is currently considered a problem by many wired and wireless providers, especially ... more Abstract: P2P is currently considered a problem by many wired and wireless providers, especially because of the large amount of traffic it generates. However, given new technology developments such as Wi-Fi Direct, we see an opportunity for P2P in mobile settings that, rather than treating mobiles as second-class citizens, seeks to take advantage of their capabilities over stationary devices.
Abstract Fair queuing is a well-studied problem in modern computer networks. However, there remai... more Abstract Fair queuing is a well-studied problem in modern computer networks. However, there remains a gap between scheduling algorithms that have provably good performance, and those that are feasible and practical to implement in high speed routers. In this paper, we propose a novel packet scheduler called Stratified Round Robin, which has low complexity, and is amenable to a simple hardware implementation.
Abstract A new model for distributed decision making, distributed game automata, focuses on how c... more Abstract A new model for distributed decision making, distributed game automata, focuses on how communication affects the quality of decisions. The goal is to limit communication between decision makers such that overhead costs are reduced but good decisions still result. Learning automata play repeated games with payoffs quantifying the performance in a distributed application.
Uploads
Papers by Joseph Pasquale