- Dangling tuple:
- Dangling tuple = a tuple in a relation that does not join with any tuple in the other relation
Example:
- Fact:
- Dangling tuples can be omitted when we perform a join operation
- More efficient
join execution:
- Transfer only the non-dangling tuples to the "join site"
Graphically:
- The semi-join (⋉)
operation:
- Definition of
the Semi join (⋉)
operation:
R ⋉ S = πR ( R ⋈ ( πY(S) ) ) where: Y = join attribute(s) in SExample:
- Definition of
the Semi join (⋉)
operation:
- Conclussion:
- R ⋉ S = the set of non-dangling tuples in R
- How to
eliminate dangling tuples
using semi-join:
- Compute
πY(S) and
transfer the
result to
R's site:
- Compute
R ⋉ πY(S) and
transfer the
result to
S's site:
- Perform
subset(R) ⋈ S
at
S's site:
- Compute
πY(S) and
transfer the
result to
R's site:
- Problem:
- We incur a
communication cost to
transfer:
- πC(S)
Graphically:
- We incur a
communication cost to
transfer:
- Solution:
- We can use bloom filters to remove most of the dangling tuples with little additional communication overhead