|
PureBytes Links
Trading Reference Links
|
same problem.
98/350/256 ( just added 128 ), 5G, can barely handle
40 charts and ATCC.
i tried it on 80 5 min / 10 day / med. complexity systems.
too slow, most of the time it just dies.
*** i too think the bottleneck is in ATCC. my systems
generate buy and sell stops, so on every bar every
5 minutes it generates an order that sends and alert into
the tracking control center. so 80 5 min charts times
about 50 bars per day, that's about 4000 alert messages
times 4 ( open, active, filled, canceled ) that's up to
16000 alert messages per day possible. with 40 charts it could
be up to 8000. so ATCC has a limit of 500 for
alerts, filled and canceled. still that's where i think the bottle neck
is. machine just gets flooded with those alert window boxes and
can not process those on time, eventually bogges down.
*** partial solution. i turned off signal alerts for active, filled
and canceled orders and only left open position and regular alerts.
so now istead of clicking off the new order alerts on every bar
pro only comes up and one alert per new signal ( new open position )
per chart...
then i set max alert settings for filled and canceled to 1 ( i don't really
care about those ). now i get only one alert per chart per signal.
so that's 80 charts times about 4 signals per day on average =
300 alerts. and only open positions get logged into the ATCC.
i am going to try full 100 charts tomorrow. but i think in fast
market it won't hold up.
*** full solution: get 1.2 Athlon AMD 200mhz bus or 1G PIII.
make sure video cards are at least 16M. and make sure
at least 256M of ram... or 384 better. then you can possibly run
100+ charts with 10-30 days of intraday data, indicators and systems
and do optimization runs at same time.
also depends on how complex your systems are, how much data
you got loaded, time frame you trade, computer speed, ram, type
of orders system generates ( market or stop ), how many charts,
any indicators loaded that updata on every tick, etc. all of those
are variable.
what i will probably end up doing is getting 1G machine to
run the system on min data length required and that's it.
just pro, systems, charts.
i will do tweaking and research/coding on a different machine in
offline mode.
just like i always did with 4.0 and 2000i.
one machine only for system deployment and trading, another
one for research, coding and tweaking. the best way to do it.
i am suprised that your 700 dually can not handle 20 charts...
wow. try turning off some the alerts like i pointed out and
limit data lenght to maybe 10 days at most.
bilo.
ps. i got now 10 workspaces, 8 charts per workspace and one
workspace with a quote screen and atcc, 5 min charts, 10 days.
when all working system resourses go down to about 20%.
----- Original Message -----
From: John A. Edeleanu <edeleanu@xxxxxxxxxxx>
To: 'Lucky Bastard' <ronin@xxxxxxxxxxx>
Cc: <omega-list@xxxxxxxxxx>
Sent: Tuesday, January 16, 2001 4:18 PM
Subject: Tradestation.com hardware
> If you want to use the tracking center for realtime signals, you will
> need a dual processor system with the fastest hard drive you can find.
> My dual 700/ 7200rpm IDE can't handle 20 charts (at open or when market
> is moderately busy) with simple systems if I send orders to the tracking
> center.
>
> If I turn off the systems, piece of cake. So, if you use lots of charts
> and only a few signals, a lesser machine will do.
>
> Regards,
>
> John
>
> PS--The next version of TSpro may do away with tracking center and
> substitute an account manager. This will supposedly cure the
> bottleneck.
>
|