1. Parallel Computation of Skyline Queries Verification COSC6490A Fall 2007 Slawomir Kmiec

2. Presentation Outline  Skyline Concepts  The Parallel Algorithm  JPF Experience  JPF Issues  Abstraction  Results  Future Work  Summary  Questions

3. Skyline Concepts In a set of points (or records) identify points that are better than (i.e. not worse than) any of the others by a given set of their attributes. NameRatingAvg. Price Parthenon5$45.00 Olympus4$40.00 Coliseum4$30.00 Pyramid3$25.00 Bombay5$35.00 Paris5$40.00 Roma4$35.00 Palermo3$30.00 Point p a is said to dominate point p b if for all i such that 1 ≤ i ≤ d we have xi(p a ) ≤ xi(p b ), and at least one of those inequalities is strict. A point p is a skyline point if it is not dominated by any other point in S. The skyline of S is denoted sky(S).

4. The Parallel Algorithm (A)  Principles: → data divided equally and distributed → local skyline is computed at each peer → size of the local skyline is shared with peers → if combined results fit on any processor → local skylines are exchanged with peers then → processor p i picks i th chunk of the combined skyline and eliminates points in it that the combined skyline dominates → local results are sent to the central process → end // of processing

5. The Parallel Algorithm (A cont.)

6. The Parallel Algorithm (B)  Principles (continued) → else // combined results do not fit on some p i → loop until required number of results is available or all p i have finished do → each processor p i picks a random set of points (in proportion of his local skyline) → this set is submitted to all peers that mark point that they dominate and marked points are returned to sender → each processor p i collects back points submitted to peers and removes marked ones from the original set but sends the remaining ones to the central processor → end loop → end // of processing

7. The Parallel Algorithm (B cont.)

8. JPF Experience  getting JPF  getting JPF to run  the Eclipse way  the Linux way  incremental examples  configuration options  JPF value-added services

9. JPF Issues  independent processors - restricted to threads  eliminate native code classes - no Swing, Sockets, NIO, Regex (Eclipse) - out of 15 just java.util.ArrayList left - eliminate Socket-oriented developed classes  search-state-space reduction - input: 10 points - 2 worker threads - operation abstraction - output discarded

10. Abstraction 2 types of developed classes left SkylineMain and SkylineWorker - workflow classes “Handler” classes - request handling classes SkylineMain SkylineMainListener SkylineMainHandler Thread Socket ServeSocket SkylineWorker SkylineWorkerListener SkylineWorkerHandler Thread Socket ServerSocket

11. Abstraction (cont.) high volume of work: - due to a lot of original code removed all GUI: - remove Swing and AWT elements asynchronous Socket messaging done as: - keep references to workers instead of addresses - eliminate the “Listener” classes - each message done as an instance of the handler - create a handler for the destination worker - execute synchronous (blocking) part of data sending - start handler to execute asynchronous processing - each type of messages split into synch- and asynch- part file IO done as: - store parameters as static constants - store input data as an array - replace input scanning with referencing the array - display or discard output String.split() method (Regex) done as: - re-done as a String manipulation method

12. Results issues reported - different issues at different settings - large volume of output to be analyzed uncaught-exception conditions - issues regarding un-synchronized access - the above as IllegalMonitorStateException dead-lock conditions - issues regarding termination conditions PreciseRaceDetector -“Unprotected Variable Access” severe warnings possibly more - it ran for a long time with no other errors - it did not finish in the time given

13. Future Work atomize code - wrap code fragments into atomic operations protect shared variable access - use locks of synchronized blocks - re-run PreciseRaceDetector run it for an extended period of time - to search the complete state space analyze the applicability of issues found - wrt the applicability to the original app - not as a result of the abstraction or transformation reduce shared data interaction - handlers to create private data structures to be quickly accepted by corresponding main process - this will allow greater robustness and redundancy

14. Summary JPF is a flexible and complex tool JPF is memory- and time- intensive JPF is a valuable verification tool the application had to be changed extensively to work with JPF potential issues were found by JPF verification = value-added service extra testing code refinement (robustness)

15. Questions???