Communication Optimization System For Dis-tributed NoSQL Data Management In Virtualiza-tion Environment

Since the rise of cloud computing and virtualization, mass data storage in the virtual-ization is much concerned by a growing number of people. Cassandra is ahigh-performance distributed NoSQL data storage system, but it faces communicationproblems under virtualization environment. For read request, reading efficiency from vir-tual machine on the same physical machine is much higher than virtual machine on dif-ferent physical machines; for write request, it must write replicas on virtual machines ofdifferent physical machines, avoiding the inaccessibility when any physical machine is outof rage.Communication optimization system for distributed NoSQL data management undervirtualization environment works like this, firstly it uses virtual machine location detection,in order to avoid single master failure problem, it uses P2P Gossip protocol, and eachphysical machine gets the virtual machine list on its own, and then disseminates the datato all other nodes, so each physical node saves all the virtual machine location information.Secondly it uses thrift to publish all the virtual machine location, and reorganize the data,so it is easy to get the location information. Thirdly it uses write node selection algorithm,when write request comes, it will select the node not in the same physical machine as thebackup node, improving the accessibility by avoiding any physical machine’s downtimeinaccessibility. Lastly it uses best reading node selection, when read request comes, it willuse the modified node selection algorithm, primarily to choose the nearest node to read,thereby improving communication efficiency.Functional testing proved that the optimization system can correctly choose a virtualmachine node for backup, any physical node goes down, it can still read what it havewritten, greatly improving the accessibility of the data. The performance testing provedthe TPS of the optimized system improved by23%.