Motivated by modern parallel computing applications, we consider the problem of scheduling parallel-task jobs with heterogeneous resource requirements in a cluster of machines. Each job consists of a set of tasks that can be processed in parallel, however, the job is considered completed only when all its tasks finish their processing, which we refer to as "synchronization" constraint. Further, assignment of tasks to machines is subject to "placement" constraints, i.e., each task can be processed only on a subset of machines, and processing times can also be machine dependent. Once a task is scheduled on a machine, it requires a certain amount of resource from that machine for the duration of its processing. A machine can process ("pack") multiple tasks at the same time, however the cumulative resource requirement of the tasks should not exceed the machine's capacity.
Our objective is to minimize the weighted average of the jobs' completion times. The problem, subject to synchronization, packing and placement constraints, is NP-hard, and prior theoretical results only concern much simpler models. For the case that migration of tasks among the placement-feasible machines is allowed, we propose a preemptive algorithm with an approximation ratio of $(6+ε)$. In the special case that only one machine can process each task, we design an algorithm with improved approximation ratio of $4$. Finally, in the case that migrations (and preemptions) are not allowed, we design an algorithm with an approximation ratio of $24$. Our algorithms use a combination of linear program relaxation and greedy packing techniques. We present extensive simulation results, using a real traffic trace, that demonstrate that our algorithms yield significant gains over the prior approaches.
HPC systems keep growing in size to meet the ever-increasing demand for performance and computational resources. Apart from increased performance, large scale systems face two challenges that hinder further growth: energy efficiency and resiliency. At the same time, applications seeking increased performance rely on advanced parallelism for exploiting system resources, which leads to increased pressure on system interconnects. At large system scales, increased communication locality can be beneficial both in terms of application performance and energy consumption. Towards this direction, sever…
Efficient scheduling of distributed deep learning (DL) jobs in large GPU clusters is crucial for resource efficiency and job performance. While server sharing among jobs improves resource utilization, interference among co-located DL jobs occurs due to resource contention. Interference-aware job placement has been studied, with white-box approaches based on explicit interference modeling and black-box schedulers with reinforcement learning. In today's clusters containing thousands of GPU servers, running a single scheduler to manage all arrival jobs in a timely and effective manner is challeng…
Today's clusters often have to divide resources among a diverse set of jobs. These jobs are heterogeneous both in execution time and in their rate of arrival. Execution time heterogeneity has lead to the development of hybrid schedulers that can schedule both short and long jobs to ensure good task placement. However, arrival rate heterogeneity, or burstiness, remains a problem in existing schedulers. These hybrid schedulers manage resources on statically provisioned cluster, which can quickly be overwhelmed by bursts in the number of arriving jobs. In this paper we propose CloudCoaster, a hyb…
We consider a distributed server system consisting of a large number of servers, each with limited capacity on multiple resources (CPU, memory, disk, etc.). Jobs with different rewards arrive over time and require certain amounts of resources for the duration of their service. When a job arrives, the system must decide whether to admit it or reject it, and if admitted, in which server to schedule the job. The objective is to maximize the expected total reward received by the system. This problem is motivated by control of cloud computing clusters, in which, jobs are requests for Virtual Machin…
Adobe Research（奥多比研究院）；IIT Kharagpur（印度理工学院卡哈拉格普尔分校）；IIT Bombay（印度理工学院孟买分校）；IIT Roorkee（印度理工学院鲁尔基分校）；Oakridge National Lab（奥克里奇国家实验室）
Large multi-tenant production clusters often have to handle a variety of jobs and applications with a variety of complex resource usage characteristics. It is non-trivial and non-optimal to manually create placement rules for scheduling that would decide which applications should co-locate. In this paper, we present DeepPlace, a scheduler that learns to exploits various temporal resource usage patterns of applications using Deep Reinforcement Learning (Deep RL) to reduce resource competition across jobs running in the same machine while at the same time optimizing for overall cluster utilizati…
We are interested in the optimal scheduling of a collection of multi-component application jobs in an edge computing system that consists of geo-distributed edge computing nodes connected through a wide area network. The scheduling and placement of application jobs in an edge system is challenging due to the interdependence of multiple components of each job, and the communication delays between the geographically distributed data sources and edge nodes and their dynamic availability. In this paper we explore the feasibility of applying Deep Reinforcement Learning (DRL) based design to address…
Resource selection and task placement for distributed execution poses conceptual and implementation difficulties. Although resource selection and task placement are at the core of many tools and workflow systems, the methods are ad hoc rather than being based on models. Consequently, partial and non-interoperable implementations proliferate. We address both the conceptual and implementation difficulties by experimentally characterizing diverse modalities of resource selection and task placement. We compare the architectures and capabilities of two systems: the AIMES middleware and Swift workfl…
As emerging deep neural network (DNN) models continue to grow in size, using large GPU clusters to train DNNs is becoming an essential requirement to achieving acceptable training times. In this paper, we consider the case where future increases in cluster size will cause the global batch size that can be used to train models to reach a fundamental limit: beyond a certain point, larger global batch sizes cause sample efficiency to degrade, increasing overall time to accuracy. As a result, to achieve further improvements in training performance, we must instead consider "strong scaling" strateg…
Although public clouds still occupy the largest portion of the total cloud infrastructure, private clouds are attracting increasing interest from both industry and academia because of their better security and privacy control. According to the existing studies, the high upfront cost is among the most critical challenges associated with private clouds. To reduce cost and improve performance, virtual machine placement (VMP) methods have been extensively investigated, however, few of these methods have focused on private clouds. This paper proposes a heterogeneous and multidimensional clairvoyant…
New HPC machines are getting close to the exascale. Power consumption for those machines has been increasing, and researchers are studying ways to reduce it. A second trend is HPC machines' growing complexity, with increasing heterogeneous hardware components and different clusters architectures cooperating in the same machine. We refer to these environments with the term heterogeneous multi-cluster environments. With the aim of optimizing performance and energy consumption in these environments, this paper proposes an Energy-Aware-Multi-Cluster (EAMC) job scheduling policy. EAMC-policy is abl…
There is increasing interest in the use of HPC machines for urgent workloads to help tackle disasters as they unfold. Whilst batch queue systems are not ideal in supporting such workloads, many disadvantages can be worked around by accurately predicting when a waiting job will start to run. However there are numerous challenges in achieving such a prediction with high accuracy, not least because the queue's state can change rapidly and depend upon many factors. In this work we explore a novel machine learning approach for predicting queue wait times, hypothesising that such a model can capture…
Recent years have witnessed a rapid growth of distributed machine learning (ML) frameworks, which exploit the massive parallelism of computing clusters to expedite ML training. However, the proliferation of distributed ML frameworks also introduces many unique technical challenges in computing system design and optimization. In a networked computing cluster that supports a large number of training jobs, a key question is how to design efficient scheduling algorithms to allocate workers and parameter servers across different machines to minimize the overall training time. Toward this end, in th…
We introduce the Balsam service to manage high-throughput task scheduling and execution on supercomputing systems. Balsam allows users to populate a task database with a variety of tasks ranging from simple independent tasks to dynamic multi-task workflows. With abstractions for the local resource scheduler and MPI environment, Balsam dynamically packages tasks into ensemble jobs and manages their scheduling lifecycle. The ensembles execute in a pilot "launcher" which (i) ensures concurrent, load-balanced execution of arbitrary serial and parallel programs with heterogeneous processor requirem…
