A Proxy Service for Multi-tenant Elastic Extension Tables - Transactions on Large-Scale-Data-and Knowledge-Centered Systems XX - page 4

The Chunk Folding is a schema mapping technique that partition logical source
tables into chunks vertically [ 1 , 9 ]. These chunks are folded in different physical tables
and joined together, where a chunk of columns is partitioned into a group of columns
and each group has a chunk id [ 9 ]. Aulbach et al. (2008) perform experiments to
measure the ef

ciency of Chunk Table and Chunk Folding techniques, and they found
that Chunk Folding technique outperform the Chunk Table technique. Additionally,
they state that the performance of this technique is enhanced by mapping the most
frequently used tenants

columns of the logical schema into conventional tables, and
the majority of tenants do not use the remaining columns in the Chunk Tables.
However, the main limitation of the Chunk Folding technique is that the common
schema must be known in advance, which is not a practical solution for multi-tenant
databases. This issue also exists in Extension Tables, Pivot Tables, and Chunk Table.
The XML Table database extension technique is a combination of relational
database systems and Extensible Markup Language (XML) [ 2 , 8 , 9 ]. The extension of
XML can be provided as native XML data type, or storing the XML document in the
database as a Character Large Object (CLOB) or Binary Large Object (BLOB) [ 2 ].
XML data type facilitating the creation of database tables, columns, views, variables
and parameters, and isolating the application from relational data model [ 8 ]. This
technique satis

'

'

needs because their data can be handled without changing
original database relational schema, and XML data type can be supported by several
relational database products [ 8 , 9 ]. In contrast, this technique reduces the data access
performance using XML

es tenants

les [ 2 ], and Heng et al. (2012) state that this technique has
the highest response time, in other words, it was the slowest technique in comparison
with Private Tables, Universal Tables, Pivot Tables, Chunk Table and Chunk Folding
techniques.
In summary, although Heng et al. (2012) use the Elastic Extension Tables (EET)
name that proposed in [ 21 ], but using this name for the Salesforce storage model is
incorrect. Heng et al. cant experiments to

evaluate retrieving data from six different multi-tenant schemas used in multi-tenant
SaaS applications including Private Tables, Universal Tables, Pivot Tables, Chunk
Table, Chunk Folding, and XML Table. The results of these experiments show that
retrieving data from Universal Table is faster than from other schemas except the Private
Tables schema. Aulbach et al. (2009) conduct experiments that compare Private Tables
schema and the Universal Table (Spare Columns) schema. The results of these exper-
iments show that the Universal Table schema has the same or better performance than
the Private Tables schema when retrieving or inserting data, except when inserting a
large number of data, the Universal Table schema is slower than the Private Tables
schema. Such experimental results lead to the conclusion that the performance of the
Universal Table schema is the best out of the

' s paper [ 12 ] conducted a number of signi

ve multi-tenant schemas, as the Private
Tables schema is only suitable for a small number of tenants. Overall, the experimental
results make the Universal Table schema the optimal schema to use for a multi-tenant
database when it is compared to Pivot Tables, Chunk Table, Chunk Folding, and XML
Table. However, as noted earlier the Universal Table can be too large introducing
overhead with the number of NULL values, which the database has to handle. Ulti-
mately, this suggests that the current available multi-tenant database schemas still have
remaining challenges and issues. Based on this conclusion, we proposed in [ 21 ] EET