Installation manual

Performance and Scalability Recommended Deployment Approach

Release Notes for Release 9.1 Service Pack 3 (9.1.30151.0) Page 11

Assessment

When adding a question to a question pool using the Find from Pool or Test button, a full table

scan was run against QTI_ASI_DATA. This increased the time taken to get data for that page.

This was corrected so that the full table scan no longer occurs.

Grade Center

The Needs Grading page takes less time to load when there are thousands of attempts on the

page.

Recommended Deployment Approach

64-Bit Deployments

Release 9.1 is the first release to offer an entire application and database footprint using 64-bit

addressable memory. To take advantage of more addressable memory, Blackboard strongly

encourages clients to deploy using 64-bit operating systems (OSs) and sub-components (Java

and SQL Server or Oracle). As part of the 64-bit deployment, Blackboard encourages clients to

take advantage of larger Java Virtual Machines (JVMs) that range from 2GB to 16GB in size. The

preferred size is 4GB, but performance testing has been conducted with acceptable performance

as high as 16GB.

Customers should begin migrating from Tomcat clusters. Tomcat clustering was introduced for

scalability purposes when the Blackboard Learn architecture was 32-bit and customers wanted

the ability to increase their memory usage on a single server. With the option to virtualize on both

32-bit and 64-bit, Blackboard’s benchmarking efforts have moved away from Tomcat cluster

deployments. Customers can achieve similar performance in a virtual environment on the same

physical server with multiple virtual instances as with a bare metal configuration with many

Tomcat cluster instances. The difference in configurations is a smaller demand on the Apache or

IIS web server fronting the Tomcat instance(s). The option to deploy a 64-bit JVM with larger

heap sizes has suppressed the need for customers to run in a cluster option.

Blackboard recommends that customers consider a deployment approach consisting of larger

64-bit JVMs that are distributed across physical servers with the option to virtualize the hardware

to take advantage of the CPU and Memory capacity of these systems.

Database Availability Options

Customers should also consider database availability options such as Microsoft SQL Server

Active/Passive or Oracle RAC One Node, which is a single instance of Oracle RAC running within

one node in a cluster. Additional nodes can be added within Oracle RAC to address scalability. In

general, both offerings are more around high-availability and will have limited or no capability to

improve performance.

For More Information

For more information about the SQL Server Failover capabilities of Blackboard Learn, contact

Blackboard Consulting.

For information about running Blackboard Learn with RAC, see the Blackboard Learn Server

Administrator Guide available on

Behind the Blackboard.