Compute vs Data Grid

Gridgain offers solution of highly computational intensive applications like website analyzers, searching, indexing searched data, large file processing and complex build process.

Gridgain offers computational grid which divides task based on computation, not divides the data. But if end of day each computing unit work on data, so what I am coming to you can achieve data grid in GG by logically dividing data into phase-wise computation.
Example: Processing on large media files

• It is a three stage process
• Uncompression
• Fiddling with inner data
• Compression
• Data Grid based Solution: 
• Divides the data 
• Send data chunks to different nodes 
  
• Process all Job on one nodes
  
• Combine the result data from result data chunks
  
• Compute Grid based Solution 
• Dedicate each job to specific node
  
• Send data in sequence to these nodes 
• Collect process data from last worker node  

Few shortfalls:

• ESB - Bring a uniform way of communication between nodes and jobs running on nodes, but this can come in future release. 
• Hardware virtualization - GG does opposite, it makes multiple computing resources into one uniform entity, where virtualization makes one into many. It is better to leave some special part for specialist like VMware and Xen. 

References: 

• http://gridgain.blogspot.com/2008/07/compute-grids-vs-data-grids.html
• http://www.gridgainsystems.com/wiki/display/GG15UG/GridGain+In+A+Glance