Schema Evolution via Multi-Version Metadata in SaaS
Shengqi Wu, Zhang Shidong, Kong Lanju · Procedia Engineering · 2012
The Basic-Table combined with Extension-Table (BT&ET) layout has become the popular data storage architecture for SaaS currently. For the sake of improving the processing efficiency we improve the BT&ET layout, and migrate some tenants’ frequently accessed extension fields into the basic table based on the tenants’ constantly need on data access. With the development of cloud computing, Multi-Tenant data need to be stored in multiple data nodes. When tenant's data storage schema evolution happens, all data nodes hava to do data migration simultaneously and tenants may perceive the influence. In order to minimize the costs and negative load of schema evolution we propose the Multi-Version metadata technology. Tenant's data on each data node may contain different version metadata, and schema envolution can run asynchronously. We carry out experiments and the results figure out that the workload decreases significantly.