Research On Multi-tenant Data Storage Mechanism Based On Universal Table

Multi-tenant data storage is a Software and services (SaaS, Software as aService) in an important research content. Aimed at the characteristics of\"singleinstance multi-tenant SaaS application, people put forward many methods to solvethe problem of multi-tenant data storage, which based on relational databaseschema mapping method is the most popular, specific and subdivided into privatetable method, wide table method, Key/Value pair (Key, Value) method and hybridmethod such as storage, storage methods have their own advantages anddisadvantages and applicable scenario. Wide table storage method in order toensure that tenants of customization requirements, often need to set aside enoughcolumn, thus greatly sparse data, at the same time schema mapping is morecomplicated. Wide table method, as a kind of effective SaaS multi-tenancy datastorage solution, need further improvement, in order to get higher space utilizationand data query efficiency.For multi-tenant data storage method of the wide table in the data spaceutilization and query efficiency, put forward a kind of multiple extension based onwide table table data storage mechanism, including extended table data storagestructure and extend the table schema mapping mechanism of mode. And based onthis, advances a general multi-tenant data storage model of data query efficiencyevaluation methods, to assess more extended table model of data storage and otherstorage efficiency of data query. In this paper, the main work includes thefollowing contents:（1）More extended table data storage structure is presented. In the wide tabledata storage model based on improved, adopts the basic table combined withextended table method instead of the original single wide table multi-tenant datastorage solution. Extension table with multiple columns of different table meetdifferent custom demand, tenant extends the tenant custom data distribution to theappropriate table, minimize null values, so reduce the amount of data of sparse, improves the data storage space utilization, also reduce the influence of sparse dataon the query efficiency.（2）Put forward more than the schema mapping mechanism in the extendedtable method. The mechanism includes business data based on normal distributionmore extension table partitioning method and selection method based on extensionof table column set aside, for multi-tenant data for reasonable storage, as well asthe tenant logical data to the physical storage location map.（3）A multi-tenant storage model based on relational algebra queryperformance evaluation methods. Through the establishment of the tenant privatetable data with the physical storage data of SQL definition, converting SQLrelational algebra, relational algebra is used to estimate the query operation of I/Ocost. The multi-tenant storage mode of inquiry into the I/O performance pricecomparison, to assess the multi-tenant data storage model of query performance.Based on wide table extension table data storage mechanism on the basis ofwide table to carry on the reasonable improvement, on the premise of meet thedemand of tenant customizing, reduced the data sparse, reduces the affect of sparsedata on the data query efficiency, improves the data storage space utilization.Multi-tenant data storage model based on the relational algebra query efficiencyevaluation methods through the pre-stored tenant SQL data definition and queryrewriting algorithm, unified query rewriting operation, unified by the I/O cost forevaluating performance of the data storage mode, so as to ensure the fairness of theevaluation method. The assessment method for essentially assess more extensiontable query performance, as well as between the model and other multi-tenantsolution, regardless of the hardware or artificial master factors, can be used toquery a fair performance evaluation, multiple tenants provide reference for thechoice of data storage model, for multi-tenant SaaS applications provide help forthe selection of data model.