面向实时云计算的负载均衡技术研究

For real-time cloud computing under large-scale traffic, load balancing which can adapt to data stream service has become an important research field. Under the environment of real-time cloud computing, the scale of data has been larger, the variety of service has been more complicated. The existing load balancing technologies face the challenge of the bottleneck of expandability, which cannot meet the need of real-time capability and reliability under high-speed stream data processing. Our research focus on load balancing architecture and algorithm for real-time cloud computing platform, on which we will propose a cloud computing platform of high availability. A load balancing mechanism based on tasks and traffic is built in this platform. For load balancing based on tasks, a task scheduling model with low complexity is presented. On the basis of this model we have provided a task scheduling and resource allocation algorithm which meets dynamic scenes in cloud environment, achieving dynamic task load balance and reducing traffic among nodes at the same time. For load balancing based on traffic, we have presented a load-balanced algorithm achieving communication consistency with no golbal communication table, which achieves traffic load balance among tasks in cloud computing. The architecture and algorithm we have porposed will improve availability and load balance of real-time cloud computing,meeting the need of stream computing in complicated business. Our research has important theoretical and practical significance and the results can be widely used in real-time cloud computing, large-scale data stream processing,etc.

对于面向大流量的实时云计算，适应数据流业务的负载均衡是一项重要的研究内容。实时云环境下数据规模日益庞大、业务多样性愈加复杂，现有的负载均衡技术面临可扩展性瓶颈的挑战，无法满足高速网络流处理的实时性可靠性要求。本课题研究面向实时云计算平台的负载均衡架构和算法，在此基础上提出一种高可用性实时云计算平台。该平台内置有任务级与流量级的负载均衡机制。对于任务级负载均衡，提出一种复杂度低的实时云任务调度模型，并基于该模型，提供了满足云环境动态场景的任务调度和资源分配算法，在云任务动态均衡的同时降低了节点间的数据流量。对于流量级负载均衡，提出一种在没有全局会话表的情况下，保持会话一致性的负载均衡算法，实现了云任务数据流量之间的均衡。本课题提出的架构与算法，提高了实时云环境的可用性和均衡性，能够满足复杂业务数据流计算的要求。研究成果可以应用于实时云平台、大规模数据流处理等领域，具有重要的理论与现实意义。

负载均衡是数据流业务对实时云计算的基本要求。如何全面、准确、快速地对数据流、任务组和节点集的三级资源进行调度成为当前学术界与产业界共同面对的重要课题。本研究立足于云环境和互联网，主要围绕面向实时云计算的负载均衡关键技术开展相关研究。在分布式流处理平台三级调度问题、地域分散云环境的高效数据交换问题、云环境下基于数据内容的匹配路由问题、云服务节点之间的资源调度分配问题等四个方面取得了重要进展。同时，研发了一个基于Docker引擎的流调度平台，一个云环境下的面向数据内容的消息中间件系统。本研究已发表学术论文11篇（其中CCF列表10篇），申请专利5项。相关成果在多个国家重大工程系统中实际应用，取得了非常显著的经济价值和社会效益，获得了教育部科学技术一等奖。