Query Processing
Query processing typically involves shipping data
- e.g. reconstructing table from distributed partitions
- e.g. join on tables stored on separate sites
Aim: minimise shipping cost (since it is a networking cost)
Shipping cost becomes the "disk access cost" of DQOpt
Can still use cost-based query optimisation
- consider possible execution plans, choose cheapest
|
![[prev]](icon/prev.gif){kind=icon}
![[next]](icon/next.gif){kind=icon}