Hi Alexander, Unfortunately, I don't know how we can improve performance significantly. Some suggestions:
1. Turn off event log. 2. Futher experiments with data structure (simplification, attributes vs elements, etc ...). Ivan Shcheklein, Sedna Team On Wed, Dec 9, 2009 at 12:17 AM, Alexander Kabanov < [email protected]> wrote: > Hi Ivan > > Thank you for your suggestions. I have managed to improve the performance > by > around 30-40% using the index. I have also sent you some event log files > - two form before and two from after I switched to index-aided inserts. > > I wonder if there is anything further we can do. As you will probably be > able to > see from the event logs, the majority of insert (in terms of size) are > those > related to the Market nodes. These have the following generic structure > > <Market> > <Bids> > <LimitOrder>...</LimitOrder> > ... > <LimitOrder>...</LimitOrder> > </Bids> > <Offers> > <LimitOrder>...</LimitOrder> > ... > <LimitOrder>...</LimitOrder> > </Offers> > </Market> > > An example of a limit order node is provided below > > <LimitOrder> > <depth>0</depth> > <Order id="10703" > > <price>91.570000000000007</price> > <quantity>226</quantity> > <is_bid>0</is_bid> > <is_market>0</is_market> > <order_time>9201</order_time> > <owner_id>12</owner_id> > </Order> > </LimitOrder> > > Now, quite often the number of LimitOrder nodes in a market node would be > large > (above 100). I was wondering if there was some way to optimise those > inserts > > > > ------------------------------------------------------------------------------ > Return on Information: > Google Enterprise Search pays you back > Get the facts. > http://p.sf.net/sfu/google-dev2dev > _______________________________________________ > Sedna-discussion mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/sedna-discussion >
------------------------------------------------------------------------------ Return on Information: Google Enterprise Search pays you back Get the facts. http://p.sf.net/sfu/google-dev2dev
_______________________________________________ Sedna-discussion mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/sedna-discussion
