Distributed computing

[lee40002]

Hi folks,

I have a bunch of unused computers in the lab that are connected to the
internet. I would like to utilize these unused computers to speed up the labview
program. So i would like to have these unused computers to share some of the
labview computing work. How do i do that??

Lee

[altenbach]

That really depends on what the program is doing. Can you tell us a little more about your application? Are there large parts of code that could execute in parallel? 

[lee40002]

Hi, My vi is designed to store the data set taken from an OSA(optical spectrum

analyzer) and plot the data using XY graph. The whole process takes about 5
seconds but I would like to shorten it to be less than 2 seconds using distributed
computing.

Lee

[altenbach]

Why is is taking 5 seconds??? What is the time limiting step?

How big is one dataset?

How many datasets need to be processed per second?

Where is the data coming from (local DAQ, instrument, etc.)?

What is "store"? Do you mean "stream to disk"?

Is the data spaced equally in X or do you really need an xy graph?

Could it be that your code is simply inefficient? For example you should make sure to operate "in place" to avoid memory thrashing. This could potentially speed things up by orders of magnitude without need of distributed computing. Can you show us some code?

[lee40002]

Hi folks,

I designed a VI that can communicate with an OSA(optical spectrum analyzer), I would like the labview to display the signal continuously in real time. The command to sweep the signal for my OSA device is “RPT”. So I have that command as my data to the GPIB write function, and then I have other GPIB read functions to retrieve the data from the local OSA.  And each data set is about 50000 bytes. I am using a wavechart function because my x dimension is spaced evenly.

[altenbach]

Your last post does not contain any questios, so I assume you are still thinking
About distibuted processing.

This makes little sense, since it seems like the displaying is the time limited step.
This necessarily needs to occur on a single node.

[CLSmith]

Lee,

Can you post your VI so that we have a better idea of exactly how you are acquiring the data and how you are processing it after acquisition?

Thanks!

Chris S.

[lee40002]

Hi,

here is my VI.

 Lee

[altenbach]

So what is the question? 

[johnsold]

While waiting for your answer to Altenbach's question, I am going to speculate: Since you originally asked about speeding up, I looked at what could be slowing down your VI.

I see two things.  One is that you are reading a maximum of 10000 bytes each iteration of the loop.  You say that a complete dataset contains 50000 points. You do not do anything to keep earlier points while collecting later ones.  This is does not slow things but is not what you said you wanted to do.  Your graph is 220 pixels wide.  If you send more than 220 points to the graph, LV will combine and reduce that to the number of pixels which will fit into the graph.  This process will not take seconds but may be slower than doing your own data reduction.  Probably insignificant with only about 1000 points.

The other thing is that the instrument may take most of your 5 seconds to respond.  If this is the case, nothing you do in LV will speed things up.  Check the manual for your instrument to see how long it takes to acquire a scan and to transmit the data.

Lynn