[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fwd: Re: RAM Disk or RAM-based Explorations



PureBytes Links

Trading Reference Links

I just caught up with this thread on RAM Drives and wanted
to post some results of tests I had run on my own system.

Please excuse my vague references to other software, I run
several EOD packages, and made these tests on another one.
I believe the results should be comparable.  I'm assuming
that the packages behave similarly in their read/write,
processor, and buffering algorithms.

------------------

About a week ago, I posted a question about other users'
experience with dual
processor systems to speed up backtesting.  Several of you
gave some excellent
advice, and I thought it might be interesting reading to
post a follow up.

For purposes of the testing, I ran the backtest on a file
that I was working
on, as well as the daily reports for the day I started
testing various system
options.  I also did not update my tickers for three days
during the testing, so the
database remained the same.  For the backtest, I used the
same profit protect
settings, time frame, and list.

My system is a homebuilt AMD K-2 450MHz, 100MHz bus,12 GB of
drive (80%
free), and 384MB of RAM.

Before the changes, the daily reports took 6:50 to
generate.  The EDS backtest
took 1:38 to run against the NASDAQ100 list.

I first tried several RAMDrive tests.  The best result I
found with a McAfee software
product called Hurricane 98.  I allocated 200Meg of my RAM
to the virtual drive.
The results were similar for case one where I loaded my
entire WinTes32 folder into
the Virtual Drive and ran the tests standalone and for case
two where I let the
software manage my RAMDrive space for process writing and
reading.  For this, I
saw an improvement of 7 seconds in my Report generation and
16 seconds in the
backtest.

I removed all of the RAMDrive programs and made some
hardware modifications.
My 12 Gig hard drive spins at 5400 rpm.  I switched this for
a 13.5Gig 7200 rpm
drive.  Using the new configuration, I found that the
reports ran in 6:31, a gain of
19 seconds.  The backtest took :40 to run, a gain of almost
1 minute over my
original set up!

I did not test the ATA66 feature in conjunction with this
hard drive upgrade.  This
is likely to increase the speed as well.  However, three
days was long enough...it
was time to get back to trading and my downloads.  :-)

Combining the RAMDrive software with the faster hard drive
produced no increase
in performance over the hard drive configuration alone.  In
fact, the reports were
slowed back down to their original 6:50.  The backtest still
ran in :40.

I suspect that the reports are more processor intensive than
the backtest, since I
saw the least improvement.

Thanks again.  In a nutshell, with your advice and $135 (an
IBM 7200rpm, 13.5
Gig, 1.9Kbyte buffer), I found a cheap way to speed up my
work.  Now, I also have
a ridiculous amount of storage space with a 12 Gig slave and
13.5 Gig master
installed...

Dave Nadeau

(If anyone is interested, please let me know.  I'll be
testing the ATA66 vs. E-IDE and
probably drop in a faster processor as prices come down).

--------------------