We are running a Dual Xeon x64 win10 box with 128GIG usable installed RAM
Our problem is around clustering a table with 350000 cases with 1000 variables [we have larger tables]. No issues naturally with a procedure like k-means but the problem is with Hierarchical clustering. Even with only 20 variables the process fails issuing an "Insufficient memory for cluster".
We had changed the workspace to 1073741824 and even 1099511627776 but still the procedure fails.
I then considered the heap space and thought the Java memory management might be at issue. I then editted the JVMCFG.INI file as follows;
-Xms from 32M to 512M
-Xmx from 225M tp 16384M, and also
-Xscmx from 32M to 256M.
However even with these changes the problem still persists.
I had even tried to flip the matrix and run a PC. Even with the memory management modifications the procedures fail due to lack of memory.
Given we have 128GIG usable memory, how can we modify the setting to run procedures for a larger table?