Very Large Scale Database Cluster
There are limitations on the largest possible database that can be managed by a single SQL Server. Once the application's data requirement grows beyond that size, data partition is necessary.
Traditionally, partitioned database is referred to as "federated databases". The overall database image is represented by a set of SQL Servers, each managing a designated partition of the overall database.
Federated databases gain in performance due to reduced dataset size managed by each SQL Server. They suffer reduced availability and increased management complexity. The service status of any SQL Server directly impacts the entire system's availability.
There are at least three different ways we can apply DBx™ technology to enhance the federated databases:
- Use DBx™ gateway to deliver higher availability and performance to the most critical data partitions.
- Use DBx™ gateway to protect the entire system; and
- Use multiple DBx™ gateways to further accelerate update and read-only queries.
Shifted replication of partitioned datasets can achieve significant performance advantages for update and read-only queries while maintaining the same high availability benefits.
For example, an INSERT query can be four times faster if the billion-row table is partitioned into four subsets and hosted by four SQL Servers. Shifted replication allows any SQL Server to crash without impacting the overall system availability. In the above figure, there are four copies of the same datasets. Each SQL Server hosts 1/4 primary partition and 3/4 backup partitions for others.
