Barry Lawson

Multiple-queue backfilling scheduling with priorities and reservations for parallel systems

We describe a new, non-FCFS policy to schedule parallel jobs on systems that may be part of a computational grid. Our algorithm continuously monitors the system (i.e., intensity of incoming jobs and variability of their resource demands) and continuously adapts its scheduling parameters to sudden workload fluctuations. The proposed policy is based on backfilling which permits job rearrangement in the waiting queue. By exploiting otherwise idle processors, this rearrangement reduces fragmentation of system resources, thereby providing higher system utilization. We propose to maintain multiple job queues that effectively separate jobs according to their projected execution time. Our policy supports different job priority classes as well as job reservations, making it appropriate for scheduling jobs on parallel systems that are part of a computational grid. Detailed performance comparisons via simulation using traces from the Parallel Workload Archive indicate that the proposed policy consistently outperforms traditional scheduling approaches.

Hardening functions for large scale distributed computations

The past few years have seen the development of distributed computing platforms designed to utilize the spare processor cycles of a large number of personal computers attached to the Internet in an effort to generate levels of computing power normally achieved only with expensive supercomputers. Such large scale distributed computations running in untrusted environments raise a number of security concerns, including the potential for intentional or unintentional corruption of computations, and for participants to claim credit for computing that has not been completed. This paper presents two strategies for hardening selected applications that utilize such distributed computations. Specifically, we show that carefully seeding certain tasks with precomputed data can significantly increase resistance to cheating (claiming credit for work not computed) and incorrect results. Similar results are obtained for sequential tasks through a strategy of sharing the computation of N tasks among K>N nodes. In each case, the associated cost is significantly less than the cost of assigning tasks redundantly.

Multiple-Queue Backfilling Scheduling with Priorities and Reservations for Parallel Systems

We describe a new, non-FCFS policy to schedule parallel jobs on systems that may be part of a computationalgrid. Our algorithm continuously monitors the system (i.e., the intensity of incoming jobs and variability of their resource demands), and adapts its scheduling parameters according to workload fluctuations. The proposed policy is based on backfilling, which reduces resource fragmentation by executing jobs in an order different than their arrivalwit hout delaying certain previously submitted jobs. We maintain multiple job queues that effectively separate jobs according to their projected execution time. Our policy supports different job priorities and job reservations, making it appropriate for scheduling jobs on parallel systems that are part of a computational grid. Detailed performance comparisons via simulation using traces from the Parallel Workload Archive indicate that the proposed policy consistently outperforms traditional backfilling.

Self-adapting backfilling scheduling for parallel systems

We focus on non-FCFS job scheduling policies for parallel systems that allow jobs to backfill, i.e., to move ahead in the queue, given that they do not delay certain previously submitted jobs. Consistent with commercial schedulers that maintain multiple queues where jobs are assigned according to the user-estimated duration, we propose a self-adapting backfilling policy that maintains multiple job queues to separate short from long jobs. The proposed policy adjusts its configuration parameters by continuously monitoring the system and quickly reacting to sudden fluctuations in the workload arrival pattern and/or severe changes in resource demands. Detailed performance comparisons via simulation using actual supercomputing, traces from the parallel workload archive indicate that the proposed policy consistently outperforms traditional backfilling.

Toward an Optimal Redundancy Strategy for Distributed Computations

Volunteer distributed computations utilize spare processor cycles of personal computers that are connected to the Internet. The related computation integrity concerns are commonly addressed by assigning tasks redundantly. Aside from the additional computational costs, a significant disadvantage of redundancy is its vulnerability to colluding adversaries. This paper presents a tunable redundancy-based task distribution strategy that increases resistance to collusion while significantly decreasing the associated computational costs. Specifically, our strategy guarantees a desired cheating detection probability regardless of the number of copies of a specific task controlled by the adversary. Though not the first distribution scheme with these properties, the proposed method improves upon existing strategies in that it requires fewer computational resources. More importantly, the strategy provides a practical lower bound for the number of redundantly assigned tasks required to achieve a given detection probability

Self-adaptive scheduler parameterization via online simulation

Although thoroughly investigated, job scheduling for high-end parallel systems remains an inexact science, requiring significant experience and intuition from system administrators to properly configure batch schedulers. Production schedulers provide many parameters for their configuration, but tuning these parameters appropriately can be very difficult - their effects and interactions are often nonintuitive. In this paper, we introduce a methodology for automating the difficult process of job scheduler parameterization. Our proposed methodology is based on using past workload behavior to predict future workload, and on online simulations of a model of the actual system to provide on-the-fly suggestions to the scheduler for automated parameter adjustment. Detailed performance comparisons via simulation using actual supercomputing traces indicate that out methodology consistently outperforms other workload-aware methods for scheduler parameterization.

Using iPodLinux in an introductory OS course

This paper describes a proof of concept for introducing iPods and iPodLinux into a one-semester introductory undergraduate operating systems course. iPodLinux is a version of the Linux operating system modified to run on iPods. We added a project to our course in which the students modified the iPodLinux kernel, and we supplemented lectures by discussing specifics of the Linux implementation as they relate to general operating systems concepts. We feel the course was much improved by these additions, with no substantive omission of regular material. Student response was very enthusiastic, and we feel the new material enhanced their course experience by providing a component that was empowering and helped to further improve their knowledge and skills.

Introducing computer science in an integrated science course

This paper describes our implementation and experience of incorporating computer science concepts into a team-taught, first-year interdisciplinary course for prospective science majors at the University of Richmond. The course integrates essential concepts from each of five STEM disciplines: biology, chemistry, computer science, mathematics, and physics. Including computer science in this course faces three primary challenges: few of the students have any CS background; the time devoted to CS instruction is reduced compared to a traditional introductory CS course; and the spirit of the course requires the CS material to be highly integrated with the other disciplines. Here we discuss our experience from three-plus years of offering the course and its impact on the major/minor pool of students in our own discipline.

Monte Carlo and discrete-event simulations in C and R

The Monte Carlo and discrete-event simulation code associated with the Simulation 101 pre-conference workshop (offered at the 2006, 2007, and 2008 Winter Simulation Conferences) is available in both C and R. This paper begins with general instructions for downloading, compiling, and executing the software. This is followed by detailed explanations of two programs that are representative of the software suite: craps uses Monte Carlo simulation to estimate the probability of winning the dice game Craps, and ssq2 uses discrete-event simulation to estimate several measures of performance associated with a single-server queue.

Evaluating Infection Prevention Strategies in Out-Patient Dialysis Units Using Agent-Based Modeling

Patients receiving chronic hemodialysis (CHD) are among the most vulnerable to infections caused by multidrug-resistant organisms (MDRO), which are associated with high rates of morbidity and mortality. Current guidelines to reduce transmission of MDRO in the out-patient dialysis unit are targeted at patients considered to be high-risk for transmitting these organisms: those with infected skin wounds not contained by a dressing, or those with fecal incontinence or uncontrolled diarrhea. Here, we hypothesize that targeting patients receiving antimicrobial treatment would more effectively reduce transmission and acquisition of MDRO. We also hypothesize that environmental contamination plays a role in the dissemination of MDRO in the dialysis unit. To address our hypotheses, we built an agent-based model to simulate different treatment strategies in a dialysis unit. Our results suggest that reducing antimicrobial treatment, either by reducing the number of patients receiving treatment or by reducing the duration of the treatment, markedly reduces overall colonization rates and also the levels of environmental contamination in the dialysis unit. Our results also suggest that improving the environmental decontamination efficacy between patient dialysis treatments is an effective method for reducing colonization and contamination rates. These findings have important implications for the development and implementation of future infection prevention strategies.