OpenTP1 Version 7 TP1/Client User's Guide TP1/Client/W, TP1/Client/P
When CUPs request service at the same time, inquires will become concentrated on the name service of a single gateway TP1/Server, thus increasing the load. If TP1/Client can select multiple TP1/Servers at random as gateways, the load on the gateway TP1/Server can be distributed.
To select gateway TP1/Servers at random, specify Y in the DCHOSTSELECT client environment definition.
When a CUP requests service, the CUP inquires the node that has the desired service from the name service of the gateway TP1/Server which is specified in the DCHOST client environment definition.
When multiple gateway TP1/Servers are specified, TP1/Client first inquires of the name service of the gateway TP1/Server that is specified at the beginning. If the service request is not accepted since TP1/Server specified at the beginning is not started or for other reasons, an attempt is made to switch to the next TP1/Server which is specified as the next gateway. This is the only case where a TP1/Server switch is attempted.
TP1/Server distributes the service requests to multiple nodes to distribute the load on the nodes. The schedule service of a node that receives service requests may sometimes transfer the service requests to TP1/Server on a different node to distribute the load. However, if TP1/Client has selected gateway TP1/Servers at random, the service requests are distributed from the already-selected gateway TP1/Server to a different TP1/Server in a different node. This may cause some overhead.
To prevent the above situation, you can distribute the load on the gateway TP1/Server selected by TP1/Client with priority. To distribute the load on the gateway TP1/Server with priority, specify Y in the DCSCDLOADPRIORITY client environment definition.
This definition is valid only when you execute RPCs without using the name service (specify Y in the DCSCDDIRECT client environment definition).
All Rights Reserved. Copyright (C) 2006, 2009, Hitachi, Ltd.