Query Plan

• One for each back-end node

• Sub-image to be filled

• I/O

  • what data blocks to read from local disks

• Communication

  • where to send each retrieved data block

• Computation

  • which (how many) data blocks to process