I am trying to evaluate jbpm and to built a stress test. I am saving a task and I am getting many database accesses in JBPM_VARIABLEINSTANCE, one for each task variable. Wouldn't it be more performant to have all n variables that a task is using in one row, instead of using n-rows? Apart from the n-updates, I guess there will be n-searches to JBPM_VARIABLEINSTANCE also when one is searching for a process instance according to the values of n variables.
I see that variables of finished processes are not purged. Will this not create a huge table JBPM_VARIABLEINSTANCE.