We usually teach our children not to be greedy. Many of the ethical principles ingrained in cultures worldwide emphasize the importance of one’s place in a larger community. However, there is need for some greed to prevent needless self-sacrifice. We explore this balancing act between caring for the broader community and caring for one’s own needs and desires through the lens of artificial decisionmaking agents. Using our modern reimplementation of the societal simulation Sugarscape, we explore the outcomes of societies where the agents employ varying amounts of greed. We show that entirely selfless and entirely selfish societies are doomed to collapse in most instances. We also show that societies that are 10% to 80% greedy have high wellbeing across a variety of commonsense metrics. These results highlight the exciting promise of developing computationally-derived guidelines for ethical behavior for individual agents which correspond to significant societal impacts.

High-throughput computing (HTC) workloads seek to complete as many jobs as possible over a long period of time. Such workloads require efficient execution of many parallel jobs and can occupy a large number of resources for a long time. As a result, full utilization is the normal state of an HTC facility. The widespread use of container orchestrators eases the deployment of HTC frameworks across different platforms, which also provides an opportunity to scale up HTC workloads with almost infinite resources on the public cloud. However, the autoscaling mechanisms of container orchestrators are primarily designed to support latency-sensitive microservices, and result in unexpected behavior when presented with HTC workloads. In this paper, we design a feedback autoscaler, High Throughput Autoscaler (HTA), that leverages the unique characteristics of the HTC workload to autoscales the resource pools used by HTC workloads on container orchestrators. HTA takes into account a reference input, the real-time status of the jobs' queue, as well as two feedback inputs, resource consumption of jobs, and the resource initialization time of the container orchestrator. We implement HTA using the Makeflow workload manager, Work Queue job scheduler, and the Kubernetes cluster manager. We evaluate its performance on both CPU-bound and IO-bound workloads. The evaluation results show that, by using HTA, we improve resource utilization by 5.6× with a slight increase in execution time (about 15%) for a CPU-bound workload, and shorten the workload execution time by up to 3.65× for an IO-bound workload.

We share our work-in-progress experiences of integrating knowledge of computing with ethics and societal impact of computing through innovative practices in the context of three different courses - an Agile software engineering project based course, a senior level, compulsory computing ethics course, and a computing service projects course. In response to the growing, urgent need for computing technology to be more responsibly and ethically developed, we believe that in addition to the current industry, the necessary change has to also be addressed at the level of computing education to more deliberately and consciously promote computing ethics. Our approach is to have students grapple with, examine and take ownership of the ethical and societal impact of the technologies they learn and build in the courses where they learn those technologies, just as they feel ownership of the technology. We also want the students to examine their perceptions of computing and ethics deeply and through multiple perspectives. We employ practices like role playing, the use of ethical frameworks, case studies, decision making game development, ethics scenario development, workshop on ethics and service learning. Our eventual goal is to have more ethics driven discussion and exercises across different parts of the computing curriculum.

Scientific workflows are typically expressed as a graph of logical tasks, each one representing a single program along with its input and output files. This poster introduces JX (JSON eXtended), a declarative language that can express complex workloads as an assembly of sub-graphs that can be partitioned in flexible ways. We present a case study of using JX to represent complex workflows for the Lifemapper biodiversity project. We evaluate partitioning approaches across several computing environments, including ND-Condor, IU-Jetstream, and SDSC-Comet, and show that a coarse partitioning results in faster turnaround times, reduced data transfer, and lower master utilization across all three systems.

Software installation remains a challenge in scientific computing. End users require custom software stacks that are not provided through commodity channels. The resulting effort needed to install software delays research in the first place, and creates friction for moving applications to new resources and new users. Ideally, end-users should be able to manage their own software stacks without requiring administrator privileges. To that end, we describe vc3-builder, a tool for deploying software environments automatically on clusters. Its primary application comes in cloud and opportunistic computing, where deployment must be performed in batch as a side effect of job execution. vc3-builder uses workflow technologies as a means of exploiting cluster resources for building in a portable way. We demonstrate the use of vc3-builder on three applications with complex dependencies: MAKER, Octave, and CVMFS, building and running on three different cluster facilities in sequential, parallel, and distributed modes.

Computer vision is experiencing an AI renaissance, in which machine learning models are expediting important breakthroughs in academic research and commercial applications. Effectively training these models, however, is not trivial due in part to hyperparameters: user-configured values that control a model's ability to learn from data. Existing hyperparameter optimization methods are highly parallel but make no effort to balance the search across heterogeneous hardware or to prioritize searching high-impact spaces. In this paper, we introduce a framework for massively Scalable Hardware-Aware Distributed Hyperparameter Optimization (SHADHO). Our framework calculates the relative complexity of each search space and monitors performance on the learning task over all trials. These metrics are then used as heuristics to assign hyperparameters to distributed workers based on their hardware. We first demonstrate that our framework achieves double the throughput of a standard distributed hyperparameter optimization framework by optimizing SVM for MNIST using 150 distributed workers. We then conduct model search with SHADHO over the course of one week using 74 GPUs across two compute clusters to optimize U-Net for a cell segmentation task, discovering 515 models that achieve a lower validation loss than standard U-Net.

Workflow management systems are widely used to express and execute highly parallel applications. For data-intensive workflows, storage can be the constraining resource: The number of tasks running at once must be artificially limited to not overflow the space available in the filesystem. It is all too easy for a user to dispatch a workflow which consumes all available storage and disrupts all system users. To address these issues, we present a three-tiered approach to workflow storage management: (1) A static analysis algorithm which analyzes the storage needs of a workflow before execution, giving a realistic prediction of success or failure. (2) An online storage management algorithm which accounts for the storage needed by future tasks to avoid deadlock at runtime. (3) A task containment system which limits storage consumption of individual tasks, enabling the strong guarantees of the static analysis and dynamic management algorithms. We demonstrate the application of these techniques on three complex workflows.

There has been a recent surge in the need for computer science educational resources. Often, this need is addressed with computer-based learning tools. However, many of these tools target coding skills and software design rather than teaching foundational topics. This work proposes an interactive tool for teaching process management using the common abstraction of the Dining Philosophers Problem. We demonstrate the effectiveness and entertainment value of this tool with feedback from undergraduate students enrolled in computer science courses. We believe this type of educational tool more adequately addresses the upcoming need for additional teaching resources in computer science, especially when used in conjunction with a traditional teaching environment.