Hi,Scott Marlowe: Following your said: 1.Can i update the postgres's update stragety to that :when update one row ,then load all table rows to memory? 2.If do that, then mean random update 's cost(time) =order update?
On Thu, Feb 21, 2008 at 3:23 PM, Scott Marlowe <[EMAIL PROTECTED]> wrote: > On Thu, Feb 21, 2008 at 1:07 AM, hewei <[EMAIL PROTECTED]> wrote: > > Hi, Scott Marlowe: > > > > You said that " As for processing them in order versus randomly,that's a > > common problem. " > > do you know why? how postgres work in this scenario. > > Pretty much the same way any database would. it's likely that your > data in the table is in some order. When you update one row, then the > next n rows are read into memory as well. Updating these is cheaper > because they don't have to be read, just flushed out to the write > ahead log. If you have very random access on a table much larger than > your shared_buffers or OS cache, then it's likely that by the time > you get back to a row on page x it's already been flushed out of the > OS or pg and has to be fetched again. >
