I have about 60 sds dataset files that I'm reading into separate sds dataset variables that I store in a dictionary (like shown below).  All together they are about 100MB in file size (*.nc files).  

    Public Class DataSetCollectionClass
        Inherits Dictionary(Of String, sds.DataSet)
    End Class

The first operation I do is go through all the files and add them to the dictionary, something like this

For each f in files
    dictionary.add(f,sds.DataSet.Open(buildURI(f))
Next

At this point, the memory use is minimal.  Then I go through the files and call GetData for each of the 8 arrays.  Each file contains about 8 1D arrays of 750 points.  This causes the memory use to increase dramatically (about 4 GB in this case).  After I have transitioned the data from the dataset to my program I would like to somehow drop the memory.  The only way I have found to do this is to purge my dictionary object of the datasets and recreate them.  So my questions are:

1.  Is there a way to release memory from the dataset variable after using GetData?  This way I don't have to purge my dictionary and recreate it.

2.  Any reason why the dataset is using so much memory?  Anyway to make the memory use more efficient?

[Edit] Oops, didn't notice how old this post was...

Do you need the dictionary for some reason? Perhaps you can just iterate through a string array of filename paths, open each DataSet referenced by the path, and call GetData there?

For example:

foreach (string f in files)
{
    using (DataSet ds = DataSet.Open(f))
    {
        //call GetData
    }
    //after the 'using' block goes out of scope,
    //the dataset is released from memory
    //...not sure what the VB.NET equivalent is
}