面向突发应急的云计算资源组织优化方法研究

Cloud computing provides a new scheme for the construction of emergency information systems. At the same time, much higher quality of services from cloud platform is requested when dealing with emergent events. Meanwhile, optimized resource organization is an effective means to improve the system performance. However, the current scheduling and resource provisioning methods have great drawbacks when dealing with sudden workload surge in emergency, especially for the surging tasks, quick resource adjustment and uncertain factors. To address these issues, this project focuses on the following four studies: 1) study of task load prediction and resource reservation methods to provide advance for cloud resource adjustment and deployment, and thus to reduce the system pressure due to the urgent resource demand in emergency; 2)study of dynamic hybrid task scheduling, breaking the hypothesis in traditional scheduling methods such that task execution time is known in prior, task execution time is certain, resources can be immediately available and the system performance is stable, etc, and controlling the accumulation of uncertain factors while scheduling to improve the timeliness and accuracy of scheduling; 3)study of rapid provisioning of cloud computing resources to avoid the phenomenon of low task completion rate due to the resource demand reservation deviation, smoothly scaling down the resource scale when tasks are reduced, avoiding the cloud system cannot timely expand when resource requirements are constantly changed; 4) study of fault-tolerant issue to improve the system reliability including the fault-tolerant technique when multiple resource or nodes fail simultaneously and dynamic backup of virtual machines to guarantee the reliable execution for tasks in emergency. By the study of this project, it can be expected to efficiently improve the processing capability of emergent tasks using cloud platform and provide key techniques and methods for the building of emergency information systems.

云计算为应急信息系统建设开辟了一种新模式，同时也对云平台的服务质量提出了更高要求。云资源组织优化是提高系统性能的有效手段，而现有方法在应对任务突发时的资源快速调整、不确定因素控制及可靠性保障等方面还存在若干瓶颈问题。本项目致力解决这些瓶颈问题，主要研究：1）任务负载预测与资源需求预约，为云资源调整部署提供提前量，缓解应急情况下资源需求紧迫性对系统造成的压力；2）任务动态调度，打破传统调度方法中任务执行时间确定、资源即时可用且性能稳定等理想性假设，控制不确定因素累加，提高调度时效性和准确性；3）云计算资源快速供给，缓解由于资源需求预约偏差带来的资源供给不足问题，在任务减少时进行资源规模平滑收缩，防止系统再扩展滞后的现象；4）可靠性保障，进行多节点同时故障容错处理及虚拟机动态热备份，保障系统可靠执行。通过本项目研究，有望提升云平台进行突发应急任务处理的能力，为应急信息系统建设提供关键技术支撑。

本项目以突发事件应急需求为牵引，采用云计算技术进行快速应急信息处理，着重对云计算资源组织优化问题进行探索性研究：1）研究了任务负载预测与资源需求预约问题，为云资源调整部署提供提前量，缓解应急情况下资源需求紧迫性对系统造成的压力，同时减少资源供给过量，提高了任务完成率和资源利用率；2）研究了面向快速响应的动态任务调度问题，打破了传统调度方法中任务集合预先已知、任务执行时间确定、资源即时可用且性能稳定等理想性假设，吸收任务调度过程中的不确定因素，控制不确定因素累加，提高了应急突发任务调度的时效性和准确性；3）研究了可快速伸缩的虚拟化云计算资源动态供给问题，避免由于资源需求预约偏差而带来的任务完成率低这一现象，在任务规模减少时使资源收缩尽量平滑，防止资源需求出现反复时，系统不能及时扩展的状况，同时保证了系统负载均衡；4）研究了面向应急云服务系统可靠性保障的容错问题，在多节点同时出错进行容错处理，提高了系统对重点关键任务的保障能力，在虚拟机层面，通过虚拟机动态热备份，实现了以较小的备份开销获得较高的系统可靠性，保证了突发应急任务的可靠执行。