The TP1/Server Base scheduler is valid as well for a service request from the CUP to an SPP. TP1/Server Base creates a schedule queue for each SPP service group and schedules a service request. If the SPP is specified as a server that receives requests from socket ('receive_from=socket' specified by the TP1/Server Base user service definition), the SPP can directly receive a service request from the CUP, not via the schedule queue.

Multiple nodes can activate the server that processes the same service group. This server is called a multi-node server. It can distribute service requests from the client to each node, decreasing the processing load. This facility is called the internode load-balancing. However, distributing requests requires extra memory for storing server information. When you are sure the single server is operating, you can maintain improved performance by not distributing service requests.

DCCLTLOADBALANCE in the client environment definition specifies whether to use the internode load-balancing.

Specify the memory size for storing server information using DCCACHE in the client definition. When you specify a larger memory size, you can store more server information, distributing service requests to many servers. If you do not use the load distributing function, allocating more memory speeds up the service information retrieval, thus improving the performance.

When you use the load distributing function, you can specify an effective period for retaining the server information in memory. The shorter the effective period is, the more often the server load information is retrieved, distributing the load based on the most recent state. Because this frequently causes an overhead for load information retrieval, however, the performance may degrade. To specify the effective period, use DCCLTCACHETIM in the client environment definition.