To balance back-end server workload based on a scenario, the prerequisites and the conditions for the target jobs described below must be satisfied.
- Prerequisites
- HiRDB/Parallel Server consisting of multiple units.
- The standby-less system switchover (effects distributed) facility is used (or can be used).
- Conditions for the target jobs
- At least two different jobs are executed by a single HiRDB.
- The back-end servers used primarily by the individual jobs (back-end servers that become overloaded) constitute only some of the available back-end servers.
- The back-end servers used primarily by the individual jobs (back-end servers that become overloaded) are mutually exclusive in terms of the processing of the individual jobs.*
- The peak periods (times) of the individual jobs do not overlap.
- The load fluctuations of each job can be predicted or anticipated.
- Transactions of the jobs that use a back-end server to be migrated terminate within short periods (at least there is no transaction that will remain uncompleted for a long time when the back-end server is to be moved).
- While a back-end server is being migrated, transactions that use that back-end server can be queued temporarily or cancelled.
- * Because the objective is to balance the load among individual units by migrating back-end servers, the individual jobs need not be completely exclusive.
- Reference note
- A back-end server is migrated when its load is high. Consequently, transactions may be placed on hold temporarily or may be cancelled, thus affecting the jobs adversely.
- The unit to which a back-end server is to be migrated may also have an unexpectedly high processing load. For this reason, the back-end server load may not be balanced even after back-end server migration.