Sites using BC4J???

Question

Hi,
I would like to know if there are any commercial sites, or any live sites existing on the web, which are developed using BC4J.
Thanks,
Amit
null

Answer 1

Steven,
Initially, we're seeing 3-10 seconds on a high end machine (PIII-866, 256MB), and near instantaneous after the first time. ( Depends upon how much you've put in your show() and how much memory you have ).
I "preserve" the commonly used frames by not disposing (uh...=null'ing.. see earlier post on this issue) them, but reuse the objects and clean things up in a show().
On LOW end client machines (PI-133, 64MB)... the frames may take 30-45 seconds ( or more ), the first time... and 5-20 seconds thereafter, depending upon the complexity.
The most complex frame has 5 tabbed panes and probably 80-100 Infoswing textfields, comboboxes, textareas and checkboxes. It also has a detail grid control, and probably 60-70 buttons. And a slew of pull down menus. It ain't my pappy's idea of a simple "frame". That single frame probably replaces 7 "frames" in our old X-Windows application.
NOTE: If you simply make frames invisible for re-use, your application WILL slow signficantly on low end machines because BC4J will try to keep all those controls on all those invisible frames up to date. This can be further aggravated by any rowMove or transaction listeners you may have implemented for the frame! So you have a trade-off on low end machines of slow re-instantiation of a frame when they need it... or fast re-instantiation but slowness for all ongoing activities!
Hmmm... I suppose I could put checks that avoids "updates" in my listeners if the frame is now shown... but is there a way to do this for the BC4J automatic stuff?
grin The users were warned a year ago, when they chose an application over JSPs (for ease of use and power of functionality.) We actually met the target I forecast... 2 minutes to startup, 30 seconds to swap frames initially, and 5-10 seconds to move between records ( Under HotSpot 1.3 ). But actually USING it under those conditions is different than agreeing to those performance expections in the design stage!
Since these are heads down multi-shift technicians/engineers who use the application continually through the day and evening, the idea is to keep the application up ( or minimized ) for the entire day or week.
We have decided to upgrade ALL our PI-133mhz/64MB machines, as they are effectively obsolete and this application pushed 'em over the edge. ( They actually have SMS, Outlook, AntiVirus, Oracle ERP apps ( GUI and Character mode ) up all at the same time on that configuration! They usually shut down Outlook or ERP to get into other apps... otherwise they went into "spin cycle" on their disk ( swapfile thrashing under W95/98) )
We're going to a min config on the client of 866mhz/128MB machines. This will be necessary anyway as we are going to Oracle Apps 11i along with a new Java based PDM system on top of my application, in the next 6-9 months. My app was just the first to expose the need for beefier PCs than Grandma uses to buy things on Ebay. On second thought, Grandma has a beefier PC so she can refresh every second to get that bid in 2 seconds before closing.
We are running into a larger than expected number of crashes of the JVMs on the client side. I'm currently testing Hotspot 1.3, Hotspot 1.3.1 (beta) and OJVM on various clients ( W95R1, W98, NT4.0SP6 and W2K ) and will report any results ( or bug patches ) I get in the near future on this board.
Some of the crashes of the Hotspot 1.3 are already documented on the java.sun.com site by other folk.
BTW - The back end server. I'm piggybacking on the Oracle ERP's HP hp-ux N-class 8 cpu server with an EMC disk array. You know, the kind they have on Galaxy class starships. My app/database fits into a tiny corner of that critter. No problems on the server side.
whew Long, rather disconnected report... we should probably start a different thread on "Client Performance" if we continue....