Re: History performance (and the usefulness of level 'none')

As we transition to the new Hyland Community, some post links may be temporarily disrupted as we finalize redirects. We apologize for the inconvenience and are working to resolve this as quickly as possible. Details.

sazzadul

Hi,

      I just wanted to share our experience with activiti which we have been running in our production environment for over an half year now with several hundred processes each day being processed and the performance is not an issue at all. Our processes are really very complex.

      We use history level activity and it caused us some trouble initially then we did a sql analysis of our database(oracle) and found the exact query which was most expensive and then added indexes on act_hi_actinst (ACT_ID_ and EXECUTION_ID_) and the problem just dissapeared.

SELECT *
  FROM (SELECT a.*, ROWNUM rnum
          FROM (SELECT *
                  FROM act_hi_actinst hai
                 WHERE hai.execution_id_ = :1
                   AND hai.act_id_ = :2
                   AND hai.end_time_ IS NULL) a
         WHERE ROWNUM < )
WHERE rnum >= :4

       Hope this would also help you in your way to evaluating activiti as process engine.

/Sazzadul

jbarrez

I'm the author of the article you reference there. Yes, history has an impact. The article proves that the overhead of the engine is really (really!) low. However, we can't remove any of the costs associated with the actual inserts and updates.

Currently, we're looking if we could handle the history asynchronously, which would solve the impact on performance. However, we're talking about release 5.12 at least here.

Also, you mention that you can do xx processes/minute. Does this mean you have only straight through processes? Most processes have wait states in between (eg user tasks). As such the life time of a process spans many hours, with only db operations every once in a while.