In data intensive apps I've had best results by pushing as much functionality as possible on to the DB server and using FDS (or WebORB) to retrieve as small a subset of data as you can get away with. If you're loading large amounts of data (I've worked with ~100k rows) then IMHO FDS or equivalent is a must, the load times are orders of magnitude quicker. There is a data management part to FDS which supports paging of data, the problem with it is that it pulls huge data sets from the DB and holds them in memory on the FDS server, which can be problematic for very large sets. We ended up creating a 'paging' system that pulls chunks of data from the DB as required and sends them back via Remote Object. Our DB people tell me that they can optimise the queries so that the DB anticipates future page requests and pre-caches them or some such... if you have a DB specialist I'd talk to them first!
Stephen. On 9 Nov 2007, at 19:03, letterpigeon wrote: > Hi all, > > We've built a flex grid that supports mass action on its data right > now (column-wise search & replace). But it just doesn't scale right > now. It can hold up to around 3K+ rows in the grid, but it's taking a > good few mins to load that up from the java backend servlet. The > requirement is for the grid to be able to support up to 100K+ or even > 1 million rows (may be not displaying them all at once, but when the > mass action/search is taken, it should be performed on the complete > dataset, not just what is being served up on the grid). > > We're just looking for ideas how this could be achieved, i.e.: if > pagination is needed, what is the best way to do? client side > pagination v.s. server? Could this be a good use case for Flex Data > Service? etc. > > Any pointers/ideas/experience would be greatly appreicated. Thanks in > advance. > > Regards,
