Under this link
http://groups.google.com/group/jbooktrader/web/JBTWebOptimizer.jar
you will find a web based version of the JBT optimizer.
It is intended to be used in headless environments with access via web
browser.
My personal use case:
I am using it in the Amazon cloud for heavy duty optimization jobs.
On my local machine some optimization jobs are running for ~8 hours.
In the Amazon cloud it is possible to use the Cluster Compute Instance
of type Quadruple Extra Large which costs $1.60 per hour.
There is an Amazon Machine Image (AMI) with the id ami-7ea24a17 with
CentOS which can be run with this instance type. The JBT benchmark
with the Equalizer strategy needs 377 seconds to complete. On Eugene’s
Intel i7-920  machine this benchmark needed 688 seconds – see this
thread for reference:

http://groups.google.com/group/jbooktrader/browse_thread/thread/cb5a20bdcfc47c81/eea0042e413c196f

So JBT in the Cloud is very fast!

How to use the JBT Web Optimizer (JBTWO):
-----------------------------------------------------------------

Start it on the command line as you start JBT. My recommended setup:
“java –server –Xmx12288M JBTWebOptimizer.jar <directory where you
start JBTWO>”
Important is the –server switch: here the Java VM runs in server mode.
In many cases this doubles the execution speed of Java programs, in
the JBT case it doubles the speed.
The Quadruple Extra Large Instance type has 23 GB RAM so 12GB for
JBTWO is OK. Of course for starters you can use –Xmx2014M
The above command line starts JBTWO. It starts an inline Jetty servlet
container. The JBTWO servlet is listening on port 80. It has no
timeout so the optimization session can run endlessly.
When testing on your local machine simply enter http://localhost in
your favorite browser. When running in the Cloud you have to enter the
public IP address of your Amazon AWS instance.
The start screen lists the strategies contained in the JAR file.
Select one strategy and press the Optimize… button. This brings you to
the optimizer screen.  You have to enter manually the path of the data
file with your tick data. In contrast to JBT no file selector is
available. You can select the data range. The calendar selector seems
to have a little display bug when it is set to status disabled. I will
have to check this. But it is only cosmetics. In the parameters
section you can edit your parameters as you are used to from JBT. When
pressing the Advanced… button you see the advanced optimization
options. As I have no slider control you have to enter manually the
parameter for the density of the Divide&Conquer optimizer.
When you have setup all your stuff you can press the Optimize button.
The optimizer starts as you know it from Swing JBT. The screen behaves
nearly exactly like his Swing based big brother.
There are some layout issues but this is only cosmetics.
There is an issue when the optimization run has a lot of results. Then
the results table becomes unresponsive, de facto you cannot use it.
This can happen when you set the number of trades to a very low
number.
In the jar file I have included the strategies delivered with JBT
7.06. If you want to test your own strategies simply put them in the /
com/jbooktrader/strategy folder in the JAR file. Technically a Java
JAR file is a ZIP file so you can do this with Winzip. In the JAR file
are many other files since I have packed it as a single Jar file, so
don’t worry.
JBTWO is based on the JBT706 code base with only one little
modifications to com.jbooktrader.platform.startup.JBookTrader.java and
com.jbooktrader.platform.util.ClassFinder.java

I will do some cleanup on my code and then publish it on this web site
in the next few days.

-- 
You received this message because you are subscribed to the Google Groups 
"JBookTrader" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/jbooktrader?hl=en.

Reply via email to