IBM Cognos Dynamic Cubes

24 IBM Cognos Dynamic Cubes

Figure 2-20 shows the availability of in-memory aggregates.

Figure 2-20 Availability of in-memory aggregates

2.2 Dynamic cubes and the Dynamic Query Mode server

Dynamic cubes are created within a DQM server. As such, an important concept to

understand is how dynamic cubes exist within a DQM server and how administrative

commands, metadata, and data requests are directed to a dynamic cube.This section

describes the interaction between a report server and a DQM server as it pertains to Cognos

Dynamic Cubes, the components of a dynamic cube within a DQM server, and finally, how

commands and requests are directed within the DQM server to a dynamic cube.

2.2.1 Dynamic cubes and report server

A dispatcher may host, at most, only a single DQM server, which appears as the

QueryService in the list of services associated with a dispatcher. All of the report servers that

run under a dispatcher communicate with the single DQM instance through a single port.

Thus, any dynamic cubes that are assigned to a dispatcher are all hosted within the single

instance of the DQM server.

It is possible to configure a cube on more than a single dispatcher, which can provide user

scalability. Note, however, that each instance of the DQM server hosts a separate version of a

cube. The interaction of a cube with the underlying relational database, and the management

of its caches, is distinct from that of any other instance of the cube on other dispatchers.

Consequently, distributing a cube, though it does allow for more users to access a single,

logical cube does incur some additional overhead:

򐂰 Each computing node to which a dynamic cube is configured must have the memory and

CPU cores necessary to support the cube and its expected user volume, taking load

balancing into account.

򐂰 If a dynamic cube is deployed across multiple nodes, the load on the underlying relational

database can be multiplied notably during the loading of in-memory aggregates if start,

restart, or data cache refresh operations are performed at the same time across the

multiple instances of the cube.

Sales Inve ntory

Sales

In ventory

Sum mary t able s availa ble

to base cube.

A base cube can construct

in-memory ag greg ate s f rom

data in fa ct t able .

A virtual cube has no direct access to

the data warehouse, so it cannot

access summary tables, or construct

in-memory aggre gat es

.