Stream-Lined Data Management

1
Stream-Lined Data Management

• Stan Zdonik
• Brown University

2
Why Streams?

• Arise whenever
• high request load could swamp the server.
• highly volatile data presents a data freshening
  problem (encourages polling).
• the storage capacity of the server is severely
  limited.
• the server is far away from the clients and
  latency really matters.
• Pervasive Computing devices often have these
  characteristics.
• e.g., sensors
• Reacting vs. lookup

3
Architectures

• What are the appropriate software components for
  handling push and pull on streams?
• Optimization
• Openness / extensibility
• What is the role of storage?

4
Additional Areas

• Languages
• Profiles
• Stream processing primitives
• Automatic data management
• Approximate queries
• Predicate-based connections