Scalability

Sybase IQ’s patented design permits databases to scale to contain many terabytes of data.

Its index-based structure allows IQ to store your data in a much smaller space than the size of the raw input data, and access it far faster than a traditional relational database. These features make Sybase IQ ideal for storing and accessing very large databases (VLDBs).

Database administrators need to understand the options and features that affect performance, and follow documented guidelines. While many default settings automatically provide the greatest efficiency, you may need to experiment with certain option settings for the fastest results, based on your configuration, your loading requirements, and your queries. Setting these options appropriately is necessary for top performance in any Sybase IQ database, but is especially important as your database grows to the multiterabyte scale.

This section introduces Sybase IQ features that help you manage a very large database, and points you to more detailed discussion and recommendations.

• Memory Use
  Allocating memory appropriately is a key factor in performance for all Sybase IQ databases.
• Data Loads
  As your database grows, it is crucial to manage data loading properly.
• Processing Threads
  Sybase IQ uses operating system threads to process queries and loads.
• Disk Space
  Learn about managing disk I/O for a Sybase IQ system.
• Intermediate Versioning
  A key aspect of managing loads and queries in larger databases is Sybase IQ’s transaction-level versioning.
• Page Size
  When you create your Sybase IQ databases, it is especially important to choose the correct IQ page size.
• Column-based Indexes
  Sybase IQ’s column-based indexing structure optimizes your ability to perform selections or calculations on attributes of interest to you.
• Query Optimizer
  The Sybase IQ query optimizer evaluates every query, choosing among various processing options to produce a query plan that offers optimal performance.
• Schema Design
  Sybase IQ often works better with denormalized schemas common in data warehouse design.
• UNION ALL Views
  Tables with a large number of rows can have lengthy load times.