Solving Operational Business Intelligence with InfoSphere Warehouse Advanced Edition

338 Solving Operational Business Intelligence with InfoSphere Warehouse Advanced Edition

9.2 Manage the cost of storage with multi-temperature

management

Managing a data warehouse with terabytes and petabytes of data requires clear

objectives followed by careful planning to ensure that all aspects of data

warehouse design are aligned. InfoSphere Warehouse provides a number of

features that are strategically aligned to the objective of effectively managing

storage costs. These include:

򐂰 Table partitioning

Often referred to as range partitioning, table partitioning allows you to

partition your table into ranges of data. A common implementation is to

partition data by date range and assign each range (data) partition to a

separate table space which can be independently managed.

For example, a fact table might be partitioned by year; all data for a given year

is placed in a single data partition. Each data partition for a table can be

located in a separate table space and each table space can be placed on

separate storage. This approach can increase performance by optimizing the

number of rows read per result rows returned because the optimizer can

eliminate entire data partitions where the predicate (where clause) specifies a

particular year. Maintenance operations such as backup can then be targeted

at active data partitions, reducing the time and cost of ownership.

򐂰 Storage groups

A storage group is a method of grouping one or more storage paths where

table data is placed. You can use storage groups to label different types of

storage according to storage speed, thus creating a multi-temperature

database.

For example, separate storage groups can be created and given storage

paths on different storage tiers: SSD storage might be assigned to a Hot

storage group, magnetic disk storage to a warm storage group, and SATA

drives to a cold storage group. Storage group functionality allows you to

asynchronously migrate a table space from one storage group to another

without affecting the availability of the data.

This approach to storage management allows you to structure your approach

to how data is aged at a (data partition) table space level. Fast storage can be

maintained for hot data only. Slower storage can support cooler data.

Using this approach you can infer that analytics queries that access cooler

data take longer to process, and operational queries that access warmer data

are more responsive.