Font Size: a A A

The Rearch On High Availability Of Master-slave Cloud Platforms

Posted on:2014-02-21Degree:MasterType:Thesis
Country:ChinaCandidate:P DengFull Text:PDF
GTID:2268330401990211Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With popularity and rapid development of cloud computing, personal computerInternet and mobile Internet are Actively move closer to the cloud computing.With theexpansive scale of cloud computing application, high availability has become one ofthe key element concerned.In this paper, by analyzing the solution of cloud computing platform in opensource community, summarizes the advantages and disadvantages of the differentsolution, base on it, raise up the ZKFHA, a solution for master-slave cloud platforms,which include five aspect design, Firstly the design of real-time backup node tomonitor the primary node and cluster state. Secondly the design of Monitor process.Thirdly the design of recovery process use Zookeeper. Fourly the design of sharestorage of meta data. Fively the design of client and DataNode retry connection. Thisfive design combine to form high availability solution of master-slave cloudplatforms.Based on the idea of ZKFHA, ZKFHA-HDFS is the solution for HDFSNameNode single point. NameNode single point is a scene of metadata saved,ZKFHA-HDFS realizes hot Standby for Namenode, uses Zookeeper to do failurerecovery, and introduce QuorumNode to share metadata. ZKFHA-MR is the solutionfor the Mapreduce Jobtracker’s availability, the single point of Jobtracker is a secne ofstateless, the ZKFHA-MR can achieve multiple backup Jobtracker, come into being aActive Jobtracker by election to service outside. And all spare Jobtracker share avirtual IP address while service outside. And realize a Recovery Mechanism to dealwith job recovery.Finally, by testing, it is proved ZKFHA-HDFS is a mechnaism with quicklyrecovery, no data lost, automatic fault recovery, stabler operation in practicalapplications and facilitate the switch compare to the other availability solutions. TheZKHFA-MR can run spare Jobtracker to take over the service while failure recovery.And the job is consistent and continuous after failover. ZKFHA can effectivelyimprove the availability of master-slave Cloud Platform.
Keywords/Search Tags:High availability, Cloud computing, ZKFHA, Hadoop, Failover
PDF Full Text Request
Related items