Grid topologies
In a grid topology, a resource manager dynamically allocates resources for job processing.
The standard InfoSphere® Information Server suite supports grid implementations on AIX® and Linux®, using the Grid Toolkit, with Platform LSF® as the resource manager. Each computer in the grid must use the same operating system and release level. The computers must be connected and accessible within the same high-speed local area network. You can set up this implementation by using the instructions in this documentation.
In an IBM® InfoSphere Information Server parallel processing topology, you modify a configuration file to assign resources to a job. To change the degree of parallelism, or the servers on which a job runs, you modify the configuration file to include the new number of nodes and associated servers. This architecture makes the association of nodes used by a job static.
With a grid implementation, a resource manager manages the pool of available resources and dynamically allocates them as jobs are submitted for processing. For each job, the resource manager waits until resources are available and then generates a configuration file that assigns resources to the job. The system minimizes the idle time for each processor and helps ensure that resources are used in the most efficient manner.
Grid topologies vary widely. They depend in part on the storage implementation such as network-attached storage (NAS) or storage area networks (SAN). However, in any grid topology, one server is designated as the conductor node, and other servers are designated as compute nodes. All jobs are initiated from the conductor node, and processed on the compute nodes.
A grid can be implemented on relatively low-cost hardware, with minimum additional costs. To scale the grid, add additional nodes and register them with the resource manager. You can take nodes out of service without halting processing of jobs within the grid.
The following diagram shows a typical grid topology. The diagram shows an IBM BladeCenter chassis with 14 blades. External disk storage is a network-attached storage (NAS) system. The storage is accessible from all blades through the NFS server installed in the first node (node 0).
In the following diagram, the metadata repository tier and services tier are installed on the first blade (node 1). The server engine, resource manager, and one parallel engine are installed on the second blade (node 2). All other blades (nodes 3 through 14) are compute nodes that provide processing power for the grid. The client tier is installed on a workstation that is connected to the blade system (this computer is not shown).
