Optimizing Sequences of DB Operations

1
Optimizing Sequences of DB Operations

• Problem
• Workflows, query generators, appl programmers all
  produce sequences of simple steps
• Op1 -gt Op2 -gt Op3 -gt OpN
• Overhead on each call
• Sometimes expensive materialization of
  intermediate results
• Goal
• Merge subsequences into more complex queries
• Op123 -gt Op45 -gt OpN
• When can you merge? How much should you merge?

2
Where to do what when in distributed query
processing

• Problem
• Data-intensive computations over distributed data
• Scientific and commercial examples
• Heterogeneous environments
• Want to spread the work over multiple systems,
  but
• Load on systems varies unpredictably
• Costs of shipping data may be prohibitive
• Limited use of and lack of mechanisms for
  function-shipping today
• Cost of evaluating more options may be
  prohibitive
• Goal
• Enhance performance of distributed queries
• Respond to changing environment