Multi-dimensional Queries in P2P Systems

1
Multi-dimensional Queries in P2P Systems
2
Applications

• Photo-sharing (photographs tagged with metadata)
• Multi-player online games (locate objects and
  other users)

3
Systems Components

• Partitioning Strategy (distribute data across
  nodes)
• Routing Strategy (transmit a query to the
  relevant nodes)

4
SCRAP SPACE-FILLING CURVES WITH RANGE
PARTITIONING
5
Partitioning Strategy

• Data are mapped into a single dimension e.g.
  (0100,0101) -gt 01000101
• Single-dimensional data are range-partitioned
  across a dynamic set of nodes

6
Routing Strategy

• Multi-dimensional range query is converted to a
  set of 1-d range queries
• Route each of 1-d range queries (Skip graph A
  Chord-like connection between the nodes, which
  promises O(logn) messages)

7
Desiderata

• Locality Nearby data points in the
  multi-dimensional space are also adjacent in the
  single dimension
• Load Balance
• Management of neighboring ranges
• Node with empty partition splits the range of
  heavily loaded node

8
Desiderata

• Minimal metadata Each node knows only partition
  boundaries of itself and its neighbors
• Low state O(logn) links per node
• Routing load balance Achieved due to the
  symmetric nature of skip graphs
• Query routing O(logn) hops for each 1-d

9
Questions?