Dynamic adaptive model of the multi-tenant data replication based on queuing theory

摘要

This Multi-tenant data for SaaS application is not available when the cluster is failure. The replication can guarantee the security of data by increasing redundancy. It also approaches to increase data availability with suitable replication policy. Now replication is in units of physical table, which doesn't fit to SaaS application in a big sharing wide table. We propose the minimum threshold of replication based on statistic probability of tenants not available, when some nodes in cluster are failure. Moreover, We present a replica adjustment strategy based on queuing theory M/M/S/∞ to respect to variation number of data request from users. The results expose SQMS is better than MQSS through simulation.