A user RDAREA should not be placed on the same disk as a system log file. When this rule is observed, input/output operations on HiRDB files that constitute system log files and on user RDAREAs can be distributed to multiple disks when a synchronization point dump is collected, thereby reducing the amount of time required for synchronization point dump processing.
A user RDAREA should not be placed on the same disk as a system RDAREA.
If you have partitioned a table by row, place the RDAREAs storing the row-partitioned table on separate back-end servers and on separate disks. The following figure shows an example of user RDAREA placement.
Figure 9-10 Example of user RDAREA placement (HiRDB/Parallel Server)
If you perform complicated query processing on tables, such as join and sort processing involving multiple back-end servers, carefully determine the placement of user RDAREAs.
If you place user RDAREAs on all back-end servers, some back-end servers' workloads become high because they not only access user RDAREAs but also execute complicated query processing. This results in reduction of overall system throughput.
If you have a sufficient number of server machines, define a back-end server that has no user RDAREA placed on it (floating server). In this way, complicated query processing is handled by the floating server, thereby reducing each back-end server's workload.