基于一致性哈希的分布式数据库性能拓展

发布时间：2018-01-10 22:04

本文关键词：基于一致性哈希的分布式数据库性能拓展　出处：《计算机应用》2016年S2期 　论文类型：期刊论文

【摘要】：当对分布式数据库进行设计开发时,通常需要考虑网络设备的异质性和系统内网络之间的可用程度和互操作程度,以及系统网络的规模和开发成本等多个因素。目前一般的分布式数据库通常被设计成一系列网络构成的多层专用网络系统,当衡量网络规模和可用程度等问题时,需要基于每个问题进行单独的设计,由于这种方案缺乏统一的解决途径,导致管理和测试的效率降低。针对传统方式引起的多次设计问题,提出一种新的一致性哈希策略来进行分布式数据库性能动态拓展的方法。首先,通过一个中心管理节点群来控制所有节点的通信过程,并完成处理故障等任务,其中每一个节点均对应一类网络管理方式及协议方式;然后,对相应的异构网络设备进行监控和统计;最后,要保证节点之间则采用统一的协议进行通信。在内网环境下对函数误差、丢失信息量和时延进行评估,测试结果表明这些指标保持稳定,与多层分布式专用网络方法相比,新的方法在节点个数增加接近8倍的情况下,平均误差仅升高了74%,丢失数据量增加了52%,往返时延在毫秒量级上基本没有变化,因此认为这种方案比传统方法能够在保持性能的前提下容纳更多节点的通信进程,表现出一定的易扩展性,且由于对异构设备兼容,使得设计及架设的代价较低。
[Abstract]:When designing and developing distributed databases, we usually need to consider the heterogeneity of network devices and the degree of availability and interoperability between networks within the system. At present, the general distributed database is usually designed as a series of network composed of multi-layer private network system. When measuring issues such as network size and availability, a separate design is needed based on each problem, as there is a lack of a unified approach to such solutions. The efficiency of management and testing is reduced. A new consistent hash strategy is proposed to dynamically extend the performance of distributed database to solve the multiple design problems caused by traditional methods. The communication process of all nodes is controlled by a group of central management nodes, and the tasks of dealing with failures are accomplished. Each node corresponds to a kind of network management mode and protocol mode. Then, the corresponding heterogeneous network equipment monitoring and statistics; Finally, the unified protocol is used to ensure the communication between nodes. In the inner network environment, the function error, the amount of lost information and the time delay are evaluated. The test results show that these indicators remain stable. Compared with the multi-layer distributed private network method, the average error of the new method increases only 74% and the lost data increases 52% when the number of nodes increases nearly 8 times. The round-trip delay has no change in millisecond order of magnitude, so it is considered that this scheme can accommodate more communication processes with more nodes than the traditional method under the premise of maintaining performance, showing a certain degree of scalability. Because of compatibility with heterogeneous devices, the cost of design and erection is lower.
【作者单位】：北京国电通网络技术有限公司;国网浙江省电力公司;清华大学电子工程系;
【基金】：国家电网公司科技项目(SGZJ0000BGJS1500433) 国网信息通信产业集团有限公司科技项目(SGITG-KJ-JSKF[2015]0003)
【分类号】：TP311.13
【正文快照】： 0引言分布式数据库具有物理分布性、逻辑整体性、站点自治性等特点,因此广泛应用于多种类型的企业和行业[1]。分布式数据库系统通常包括许多的设备和管理域,这些设备之间以及管理域之间通常具有不同的特性,由统一的分布式数据库系统进行管理。一个好的分布式数据库系统需要具

【相似文献】