I am not perfectly sure what you mean by "transactional tables". Let me use the term "data handled in transaction" or "dhit" and try to forecast the issues.
AFAIK the "dhit" at the current OpenObject is in the form of python objects. They are stored in the server's memory, handled by its garbage collector and so on. Yes, there is many SQL commands issued inside each "rpc action", which are then committed at the end of the transaction. Still, Python has a reference of nearly all objects involved there. In some experiments I've been making, the counters often show as many as 5k SQL commands ! inside one transaction. In some case, 50k and 100k too! That is a problem, for sure. As a quick remedy, the object model should be planned so that number of "dhit" objects is kept minimal (say, do you really need more than 100 lines in an invoice? , can you limit your search in products when you do the query? ). Then, Postgres is a great tool in optimizing the data handling. Just like any decent database, it is destined to hold huge datasets and produce a reasonably fast result. The key there is that you don't foolishly ask for the whole dataset, but the right part of it. Given that you have the right indexes and no excess triggers, most db transactions perform fast and don't really depend on the rowcount of the table. My experience (since 2003) in Postgres has not yet found an upper limit to the data size. Of course, we are *only* talking about using the latest Postgres in Linux (or Solaris, perhaps), using decent hardware and not compromize for anything less than those. My humble conclusion is that now you should expect moderate performance with Object model, you may be lucky enough to store quite many rows before the model does something stu^H^H^Hsuboptimal and slows down. By the time you will have stored more data, all the components (OpenObject, db, Linux, hardware) will have progressed so that you will enjoy the desired scalability. -------------------- m2f -------------------- -- http://www.openerp.com/forum/viewtopic.php?p=29184#29184 -------------------- m2f -------------------- _______________________________________________ Tinyerp-users mailing list http://tiny.be/mailman/listinfo/tinyerp-users
