An official website of the United States government
High performance computing (HPC) is undergoing significant changes. The emerging HPC applications comprise both compute- and data-intensive applications. To meet the intense I/O demand from emerging data-intensive applications, burst buffers are deployed in production systems. Existing HPC schedulers are mainly CPU-centric. The extreme heterogeneity of hardware devices, combined with workload changes, forces the schedulers to consider multiple resources (e.g., burst buffers) beyond CPUs, in decision making. In this study, we present a multi-resource scheduling scheme named BBSched that schedules user jobs based on not only their CPU requirements, but also other schedulable resources such as burst buffer. BBSched formulates the scheduling problem into a multi-objective optimization (MOO) problem and rapidly solves the problem using a multi-objective genetic algorithm. The multiple solutions generated by BBSched enables system managers to explore potential tradeoffs among various resources, and therefore obtains better utilization of all the resources. The trace-driven simulations with real system workloads demonstrate that BBSched improves scheduling performance by up to 41% compared to existing methods, indicating that explicitly optimizing multiple resources beyond CPUs is essential for HPC scheduling.
more »
« less
RLScheduler: An Automated HPC Batch Job Scheduler Using Reinforcement Learning
Today’s high-performance computing (HPC) platforms are still dominated by batch jobs. Accordingly, effective batch job scheduling is crucial to obtain high system efficiency. Existing HPC batch job schedulers typically leverage heuristic priority functions to prioritize and schedule jobs. But, once configured and deployed by the experts, such priority function scan hardly adapt to the changes of job loads, optimization goals, or system settings, potentially leading to degraded system efficiency when changes occur. To address this fundamental issue, we present RLScheduler, an automated HPC batch job scheduler built on reinforcement learning. RLScheduler relies on minimal manual interventions or expert knowledge, but can learn high-quality scheduling policies via its own continuous ‘trial and error’. We introduce a new kernel-based neural network structure and trajectory filtering mechanism in RLScheduler to improve and stabilize the learning process. Through extensive evaluations,we confirm that RLScheduler can learn high-quality scheduling policies towards various workloads and various optimization goals with relatively low computation cost. Moreover, we show that the learned models perform stably even when applied to unseen workloads, making them practical for production use.
more »
« less
- Award ID(s):
- 1817089
- PAR ID:
- 10196073
- Date Published:
- Journal Name:
- SC'20: The International Conference for High Performance Computing, Networking, Storage, and Analysis 2020
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Due to a growing interest in deep learning applications [5], compute-intensive and long-running (hours to days) training jobs have become a significant component of datacenter workloads. A large fraction of these jobs is often exploratory, with the goal of determining the best model structure (e.g., the number of layers and channels in a convolutional neural network), hyperparameters (e.g., the learning rate), and data augmentation strategies for the target application. Notably, training jobs are often terminated early if their learning metrics (e.g., training and validation accuracy) are not converging, with only a few completing successfully. For this motivating application, we consider the problem of scheduling a set of jobs that can be terminated at predetermined checkpoints with known probabilities estimated from historical data. We prove that, in order to minimize the time to complete the first K successful jobs on a single server, optimal scheduling does not require preemption (even when preemption overhead is negligible) and provide an optimal policy; advantages of this policy are quantified through simulation. Related Work. While job scheduling has been investigated extensively in many scenarios (see [6] and [2] for a survey of recent result), most policies require that the cost of waiting times of each job be known at scheduling time; in contrast, in our setting the scheduler does not know which job will be the K-th successful job, and sojourn times of subsequent jobs do not contribute to the target metric. For example, [4, 3] minimize makespan (i.e., the time to complete all jobs) for known execution times and waiting time costs; similarly, Gittins index [1] and SR rank [7] minimize expected sojourn time of all jobs, i.e., both successfully completed jobs and jobs terminated early. Unfortunately, scheduling policies not distinguishing between these two types of jobs may favor jobs where the next stage is short and leads to early termination with high probability, which is an undesirable outcome in our applications of interest.more » « less
-
Data center workloads are composed of multiresource jobs requiring a variety of computational resources including CPU cores, memory, disk space, and hardware accelerators. Mod- ern servers can run multiple jobs in parallel, but a set of jobs can only run in parallel if the server has sufficient resources to satisfy the demands of each job. It is generally hard to find sets of jobs that perfectly utilize all server resources, and choosing the wrong set of jobs can lead to low resource uti- lization. This raises the question of how to allocate resources across a stream of arriving multiresource jobs to minimize the mean response time across jobs — the mean time from when a job arrives to the system until it is complete. Current policies for scheduling multiresource jobs are com- plex to analyze and hard to implement. We propose a class of simple policies, called Markovian Service Rate (MSR) policies. We show that the class of MSR policies is throughput- optimal, in that if a policy exists that can stabilize the sys- tem, then an MSR policy exists that stabilizes the system. We derive bounds on the mean response time under an MSR policy, and show how our bounds can be used to choose an MSR policy that minimizes mean response time.more » « less
-
Grid Engine is a Distributed Resource Manager (DRM), that manages the resources of distributed systems (such as Grid, HPC, or Cloud systems) and executes designated jobs which have requested to occupy or consume those resources. Grid Engine applies scheduling policies to allocate resources for jobs while simultaneously attempting to maintain optimal utilization of all machines in the distributed system. However, due to the complexity of Grid Engine's job submission commands and complicated resource management policies, the number of faulty job submissions in data centers increases with the number of jobs being submitted. To combat the increase in faulty jobs, Grid Engine allows administrators to design and implement Job Submission Verifiers (JSV) to verify jobs before they enter into Grid Engine. In this paper, we will discuss a Job Submission Verifier that was designed and implemented for Univa Grid Engine, a commercial version of Grid Engine, and thoroughly evaluated at the High Performance Computing Center of Texas Tech University. Our newly developed JSV communicates with Univa Grid Engine (UGE) components to verify whether a submitted job should be accepted as is, or modified then accepted, or rejected due to improper requests for resources. It had a substantial positive impact on reducing the number of faulty jobs submitted to UGE by far. For instance, it corrected 28.6% of job submissions and rejected 0.3% of total jobs from September 2018 to February 2019, that may otherwise lead to long or infinite waiting time in the job queue.more » « less
-
High-performance computing (HPC) resources are used for compute-demanding calculations in various fields of science and engineering. They are large computational facilities utilized by many users simultaneously. High utilization often leads to high waiting times. Simulating users' behavior on such a system can help with future system design, develop user interventions, and ultimately improve the user’s experience and resource utilization. Here, we present HPCMod, an Agent-Based Modeling Framework for Modeling Users on HPC Resources. The key concept of the framework is the representation of the user's computational needs: the user project is represented as a collection of possibly dependent compute tasks. Each task can be executed as a single compute job or a series of jobs, depending on the task size. Some tasks can be too big to be executed in one chunk; such a situation often occurs during molecular dynamics simulation. There are multiple ways in which tasks can be split into jobs, and users will make their decisions based on previous experience, application parallel scalability, and available resources. For example, a user's compute task requires 32 node hours; it can be executed in multiple ways: a single 32-hour job on one node, two sequential 16-hour jobs on one node, one 16-hour job on two nodes, and so on. In the HPCMod, we implemented three models: 1) historical replay of compute jobs, 2) simulation of reconstituted compute tasks using historical job sizes, and 3) adaptive compute tasks splitting where users can modify jobs parameters given available resources till the execution of the next job in line. The framework was tested on a ten-node test system and a larger 1,736-node system modeled after a portion of TACC Stampede-2. The HPC resource model implements a first in first out (FIFO) scheduler with backfill scheduling. The initial results showed that on a tiny system, adaptive task-splitting is beneficial for the user but leads to a larger number of jobs. On a large system, the adaptive task-splitting was also very beneficial, decreasing waiting times for users using this strategy almost two times; however, other users got a 5% increase in their wait time. Further investigation is needed as the current task reconstitution algorithm is deterministic and does not allow quantification of job recombination uncertainties. The Julia-based implementation is fast: five years of historic workflow consisting of a million jobs and a one-hour stepping took around three minutes.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
