Resource Request Routing

 

When an object (data or instruction) referenced by a task is unknown to a resource manager, it must check to see if the object is recognized by other nodes in the system. Instead of querying all other nodes directly, the resource request is passed directly up the hierarchy mentioned earlier.

Each node recognizes at least two nodes (per process), and possible many more. The nodes are identified as :

primary

A node in higher level than the current processor. If the current node is in the top level in the hierarchy, it has no primary.

alternate

A second node in higher level than the current processor. If the current node is in the top level in the hierarchy, it has no alternate. The alternate is provided for fault tolerance.

peer

A node at the same level as the current processor.

slave

A node in a lower level, for which the current processor is the primary.

  If a node being queried doesn't have the requested data object, it forwards the query to the primary. The forwarding information ( i.e. who is the primary ?) is obtained from the local (on the machine being queried) process environment associated with the query.