Search NASA⌕ Search

Engineering topics

Lan, Zhiling

Publications and source records attributed to Lan, Zhiling.

DRAS: Deep Reinforcement Learning for Cluster Scheduling in High Performance Computing

Cluster schedulers are crucial in high-performance computing (HPC). They determine when and which user jobs should be allocated to available system resources. Existing cluster scheduling heuristics are developed by human experts based on their experience with specific HPC systems and workloads. However, the increasing complexity of computing systems and the highly dynamic nature of application workloads have placed tremendous burden on manually designed and tuned scheduling heuristics. More aggressive optimization and automation are needed for cluster scheduling in HPC. In this work, we present an automated HPC scheduling agent named DRAS (Deep Reinforcement Agent for Scheduling) by leveraging deep reinforcement learning. DRAS is built on a hierarchical neural network incorporating special HPC scheduling features such as resource reservation and backfilling. An efficient training strategy is presented to enable DRAS to rapidly learn the target environment. Once being provided a specific scheduling objective given by the system manager, DRAS automatically learns to improve its policy through interaction with the scheduling environment and dynamically adjusts its policy as workload changes. We implement DRAS into a HPC scheduling platform called CQGym. CQGym provides a common platform allowing users to flexibly evaluate DRAS and other scheduling methods such as heuristic and optimization methods. Furthermore, the experiments using CQGym with different production workloads demonstrate that DRAS outperforms the existing heuristic and optimization approaches by up to 50%.

97 MATHEMATICS AND COMPUTING↗

Performance and power modeling and prediction using MuMMI and 10 machine learning methods

Energy-efficient scientific applications require insight into how high performance computing system features impact the applications' power and performance. This insight can result from the development of performance and power models. Here, in this article, we use the modeling and prediction tool MuMMI (Multiple Metrics Modeling Infrastructure) and 10 machine learning methods to model and predict performance and power consumption and compare their prediction error rates. We use an algorithm-based fault-tolerant linear algebra code and a multilevel checkpointing fault-tolerant heat distribution code to conduct our modeling and prediction study on the Cray XC40 Theta and IBM BG/Q Mira at Argonne National Laboratory and the Intel Haswell cluster Shepard at Sandia National Laboratories. Our experimental results show that the prediction error rates in performance and power using MuMMI are less than 10% for most cases. By utilizing the models for runtime, node power, CPU power, and memory power, we identify the most significant performance counters for potential application optimizations, and we predict theoretical outcomes of the optimizations. Based on two collected datasets, we analyze and compare the prediction accuracy in performance and power consumption using MuMMI and 10 machine learning methods.

97 MATHEMATICS AND COMPUTING↗

Union: An Automatic Workload Manager for Accelerating Network Simulation

With the rapid growth of the machine learning applications, the workloads of future HPC systems are anticipated to be a mix of scientific simulation, big data analytics, and machine learning applications. Simulation is a great research vehicle to understand the performance implications of co-running scientific applications with big data and machine learning workloads on large-scale systems. In this paper, we present Union, a workload manager that provides an automatic framework to facilitate hybrid workload simulation in CODES. Furthermore, we use Union, along with CODES, to investigate various hybrid workloads composed of traditional simulation applications and emerging learning applications on two dragonfly systems. The experiment results show that both message latency and communication time are important performance metrics to evaluate network interference. Network interference on HPC applications is more reflected by the message latency variation, whereas ML application performance depends more on the communication time.

heterogeneous workloads↗

Extreme Heterogeneity 2018 - Productive Computational Science in the Era of Extreme Heterogeneity: Report for DOE ASCR Workshop on Extreme Heterogeneity

The 2018 Basic Research Needs Workshop on Extreme Heterogeneity identified five Priority Research Directions for realizing the capabilities needed to address the challenges posed in this era of rapid technological change. This report captures the outcomes of that workshop. In the context of extreme heterogeneity, it defines basic research needs and opportunities in computer science research to develop smart and trainable operating and runtime systems, programming environments, and predictive tools that will make future systems easier to adapt to scientists’ computing needs and easier for facilities to deploy securely.

97 MATHEMATICS AND COMPUTING↗

Experience and Practice of Batch Scheduling on Leadership Supercomputers at Argonne

The mission of the DOE Argonne Leadership Computing Facility (ALCF) is to accelerate major scientific discoveries and engineering breakthroughs for humanity by designing and providing world-leading computing facilities in partnership with the computational science community. The ALCF operates supercomputers that are generally amongst the Top 5 fastest machines in the world. Specifically, ALCF is looking for the science that is either too big to run anywhere else, or it would take so long as to be impractical (i.e., “capability jobs”). At ALCF, batch scheduling plays a critical role for achieving a set of site goals within a set of constraints. While system utilization is an important goal at ALCF, its largest mission constraint is to enable extreme scale parallel jobs to take precedence. In this paper, we will describe the specific scheduling goals and constraints, analyze the workload traces collected in 2013–2017 from the 48-rack petascale supercomputer Mira, and discuss the upcoming scheduling challenges at ALCF.

Allcock, William↗

Toward General Software Level Silent Data Corruption Detection for Parallel Applications

Silent data corruption (SDC) poses a great challenge for high-performance computing (HPC) applications as we move to extreme-scale systems. Mechanisms have been proposed that are able to detect SDC in HPC applications by using the peculiarities of the data (more specifically, its “smoothness” in time and space) to make predictions. However, these data-analytic solutions are still far from fully protecting applications to a level comparable with more expensive solutions such as full replication. In this work, we propose partial replication to overcome this limitation. More specifically, we have observed that not all processes of an MPI application experience the same level of data variability at exactly the same time. Thus, we can smartly choose and replicate only those processes for which the lightweight data-analytic detectors would perform poorly. In addition, we propose a new evaluation method based on the probability that a corruption will pass unnoticed by a particular detector (instead of just reporting overall single-bit precision and recall). Here in our experiments, we use four applications dealing with different explosions. Finally, our results indicate that our new approach can protect the MPI applications analyzed with 7–70% less overhead (depending on the application) than that of full duplication with similar detection recall.

97 MATHEMATICS AND COMPUTING↗

I/O-aware bandwidth allocation for petascale computing systems

In the Big Data era, the gap between the storage performance and an appli- cation's I/O requirement is increasing. I/O congestion caused by concurrent storage accesses from multiple applications is inevitable and severely harms the performance. Conventional approaches either focus on optimizing an ap- plication's access pattern individually or handle I/O requests on a low-level storage layer without any knowledge from the upper-level applications. In this paper, we present a novel I/O-aware bandwidth allocation framework to coordinate ongoing I/O requests on petascale computing systems. The motivation behind this innovation is that the resource management system has a holistic view of both the system state and jobs' activities and can dy- namically control the jobs' status or allocate resource on the y during their execution. We treat a job's I/O requests as periodical subjobs within its lifecycle and transform the I/O congestion issue into a classical scheduling problem. Based on this model, we propose a bandwidth management mech- anism as an extension to the existing scheduling system. We design several bandwidth allocation policies with different optimization objectives either on user-oriented metrics or system performance. We conduct extensive trace- based simulations using real job traces and I/O traces from a production IBM Blue Gene/Q system at Argonne National Laboratory. Experimental results demonstrate that our new design can improve job performance by more than 30%, as well as increasing system performance.

I/O congestion↗

3-Dimensional root cause diagnosis via co-analysis. In: ICAC '12 Proceedings of the 9th international conference on Autonomic computing

With the growth of system size and complexity, reliability has become a major concern for large-scale systems. Upon the occurrence of failure, system administrators typically trace the events in Reliability, Availability, and Serviceability (RAS) logs for root cause diagnosis. However, RAS log only contains limited diagnosis information. Moreover, the manual processing is time-consuming, error-prone, and not scalable. To address the problem, in this paper we present an automated root cause diagnosis mechanism for large-scale HPC systems. Our mechanism examines multiple logs to provide a 3-D fine-grained root cause analysis. Here, 3-D means that our analysis will pinpoint the failure layer, the time, and the location of the event that causes the problem.

Zheng, Ziming↗

Filtering log data: finding the needles in the haystack

Log data is an incredible asset for troubleshooting in large-scale systems. Nevertheless, due to the ever-growing system scale, the volume of such data becomes overwhelming, bringing enormous burdens on both data storage and data analysis. To address this problem, we present a 2-dimensional online filtering mechanism to remove redundant and noisy data via feature selection and instance selection. The objective of this work is two-fold: (i) to significantly reduce data volume without losing important information, and (ii) to effectively promote data analysis. We evaluate this new filtering mechanism by means of real environmental data from the production supercomputers at Oak Ridge National Laboratory and Sandia National Laboratory. Our preliminary results demonstrate that our method can reduce more than 85% disk space, thereby significantly reducing analysis time. Moreover, it also facilitates better failure prediction and diagnosis by more than 20%, as compared to the conventional predictive approach relying on RAS (Reliability, Availability, and Serviceability) events alone.

failure prediction↗

3-Dimensional Root Cause Diagnosis via Co-analysis

With the growth of system size and complexity, reliability has become a major concern for large-scale systems. Upon the occurrence of failure, system administrators typically trace the events in Reliability, Availability, and Serviceability (RAS) logs for root cause diagnosis. However, RAS log only contains limited diagnosis information. Moreover, the manual processing is time-consuming, error-prone, and not scalable. To address the problem, in this paper we present an automated root cause diagnosis mechanism for large-scale HPC systems. Our mechanism examines multiple logs to provide a 3-D fine-grained root cause analysis. Here, 3-D means that our analysis will pinpoint the failure layer, the time, and the location of the event that causes the problem. We evaluate our mechanism by means of real logs collected from a production IBM Blue Gene/P system at Oak Ridge National Laboratory. It successfully identifies failure layer information for 219 failures during 23-month period. Furthermore, it effectively identifies the triggering events with time and location information, even when the triggering events occur hundreds of hours before the resulting failures.

Zheng, Ziming↗