tracking in a spaghetti bowl monitoring transactions
play

Tracking in a Spaghetti Bowl: Monitoring Transactions Using - PowerPoint PPT Presentation

Dec 06, 2022 •207 likes •1.45k views

Tracking in a Spaghetti Bowl: Monitoring Transactions Using Footprints Anima Anandkumar 1 , 2 Chatschik Bisdikian 2 Dakshi Agrawal 2 1 ECE Dept., Cornell University, Ithaca, NY 14853. 2 Networking Tech., IBM Watson Research, Hawthorne, NY 10532.

  1. Problem Formulation ? ? ? ? ? S 2 S 3 S 1 1 ? 2 S 0 S 4 3 ? � � � � Information in Footprint State & time stamp, optionally tokens (identifiers) Real-time match: all footprints may not yet have been generated Problem Statement Given footprints and transaction model, find maximum likelihood match between footprints and transactions Questions & Issues Complexity of maximum likelihood? Decentralized implementation? Probabilistic bounds on accuracy? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 4 / 20

  2. Problem Formulation ? ? ? ? ? S 2 S 3 S 1 1 ? 2 S 0 S 4 3 ? � � � � Information in Footprint State & time stamp, optionally tokens (identifiers) Real-time match: all footprints may not yet have been generated Problem Statement Given footprints and transaction model, find maximum likelihood match between footprints and transactions Questions & Issues Complexity of maximum likelihood? Decentralized implementation? Probabilistic bounds on accuracy? Effect of transaction model ? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 4 / 20

  3. Problem Formulation Transaction Model ? ? ? ? ? S 2 S 3 S 1 1 ? 2 S 0 S 4 3 ? � � � � Information in Footprint State & time stamp, optionally tokens (identifiers) Real-time match: all footprints may not yet have been generated Problem Statement Given footprints and transaction model, find maximum likelihood match between footprints and transactions Questions & Issues Complexity of maximum likelihood? Decentralized implementation? Probabilistic bounds on accuracy? Effect of transaction model ? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 4 / 20

  4. Problem Formulation Transaction Model ? ? ? ? ? IID transitions of transactions S 2 S 3 S 1 1 ? 2 S 0 S 4 3 ? � � � � Information in Footprint State & time stamp, optionally tokens (identifiers) Real-time match: all footprints may not yet have been generated Problem Statement Given footprints and transaction model, find maximum likelihood match between footprints and transactions Questions & Issues Complexity of maximum likelihood? Decentralized implementation? Probabilistic bounds on accuracy? Effect of transaction model ? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 4 / 20

  5. Problem Formulation Transaction Model ? ? ? ? ? IID transitions of transactions S 2 S 3 S 1 1 ? 2 S 0 Acyclic semi-Markov process S 4 3 ? � � � � Information in Footprint State & time stamp, optionally tokens (identifiers) Real-time match: all footprints may not yet have been generated Problem Statement Given footprints and transaction model, find maximum likelihood match between footprints and transactions Questions & Issues Complexity of maximum likelihood? Decentralized implementation? Probabilistic bounds on accuracy? Effect of transaction model ? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 4 / 20

  6. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  7. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  8. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems Transaction 1 Transaction 2 Y 0 W (1 , 1) . W (2 , 2) . S 0 S 1 W (2 , 1) . W (1 , 2) . Y 1 Footprint Node ccdf Node Reduction to bipartite minimum weight perfect matching Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  9. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems Transaction 1 Transaction 2 Y 0 W (1 , 1) . W (2 , 2) . S 0 S 1 W (2 , 1) . W (1 , 2) . Y 1 Footprint Node ccdf Node Reduction to bipartite minimum weight perfect matching Reduction to FIFO for a class of transition-time pdfs Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  10. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems Transaction 1 Transaction 2 Y 0 W (1 , 1) . W (2 , 2) . S 0 S 1 W (2 , 1) . W (1 , 2) . Y 1 Footprint Node ccdf Node Reduction to bipartite minimum weight perfect matching Reduction to FIFO for a class of transition-time pdfs Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  11. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems Transaction 1 Transaction 2 Y 0 W (1 , 1) . W (2 , 2) . S 0 S 1 W (2 , 1) . W (1 , 2) . Y 1 Footprint Node ccdf Node Reduction to bipartite minimum weight perfect matching Reduction to FIFO for a class of transition-time pdfs Worst Case Analysis of Maximum Likelihood Performance Uniform and exponential transition times are worst-case distributions Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  12. Summary of Results Problem Statement Maximum likelihood match between footprints & transactions Maximum Likelihood for Two-state Systems Transaction 1 Transaction 2 Y 0 W (1 , 1) . W (2 , 2) . S 0 S 1 W (2 , 1) . W (1 , 2) . Y 1 Footprint Node ccdf Node Reduction to bipartite minimum weight perfect matching Reduction to FIFO for a class of transition-time pdfs Worst Case Analysis of Maximum Likelihood Performance Uniform and exponential transition times are worst-case distributions Equal to reciprocal of number of unique perfect matchings Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 5 / 20

  13. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  14. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Constructive proof of optimality: Decentralized rules Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  15. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Constructive proof of optimality: Decentralized rules Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  16. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Constructive proof of optimality: Decentralized rules 3 4 2 1 5 1 3 2 ({1,3},{2,4,5}) ({2},{3}) 4 2 5 3 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  17. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Constructive proof of optimality: Decentralized rules 3 4 2 1 5 1 3 2 ({1,3},{2,4,5}) ({2},{3}) 4 2 5 3 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  18. Summary of Results (cont.) Maximum Likelihood for Multi-state Systems Series of bipartite matchings in high-level two-state systems Constructive proof of optimality: Decentralized rules 3 4 2 1 5 1 3 2 ({1,3},{2,4,5}) ({2},{3}) 4 2 5 3 Presence of Tokens For linear model, ML rule is still decentralized bipartite matching Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 6 / 20

  19. Outline Introduction 1 Two-state System 2 Multi-state System 3 Conclusion 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 7 / 20

  20. Two-state System S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  21. Two-state System S 0 S 1 1 2 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  22. Two-state System S 0 S 1 1 2 Time S 0 ? S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  23. Two-state System S 0 S 1 1 2 Time S 0 ? S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  24. Two-state System S 0 S 1 1 2 3 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  25. Two-state System S 0 S 1 1 2 3 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  26. Two-state System S 0 S 1 1 2 3 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  27. Two-state System S 0 S 1 1 2 3 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  28. Two-state System S 0 S 1 1 2 3 Time S 0 S 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  29. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  30. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Y 0 Y 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  31. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 Y 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  32. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 Y 1 Footprint Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  33. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 Y 1 Footprint Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  34. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ f T ( Y 1 (1) − Y 0 (1))] . Y 1 Footprint Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  35. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ f T ( Y 1 (1) − Y 0 (2))] . − log[ f T ( Y 1 (1) − Y 0 (1))] . Y 1 Footprint Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  36. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ f T ( Y 1 (1) − Y 0 (2))] . − log[ f T ( Y 1 (1) − Y 0 (1))] . Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  37. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ f T ( Y 1 (1) − Y 0 (2))] . − log[ f T ( Y 1 (1) − Y 0 (1))] . Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  38. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (2))] . F T ( Y 1 (1) − Y 0 (2))] − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (1))] . F T ( Y 1 (1) − Y 0 (1))] Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  39. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (2))] . F T ( Y 1 (1) − Y 0 (2))] − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (1))] . F T ( Y 1 (1) − Y 0 (1))] Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  40. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (2))] . F T ( Y 1 (1) − Y 0 (2))] − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (1))] . F T ( Y 1 (1) − Y 0 (1))] Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  41. Two-state System S 0 S 1 1 2 3 Time Count S 0 Time S 1 Batch Transaction 1 Transaction 2 Y 0 − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (2))] . F T ( Y 1 (1) − Y 0 (2))] − log[ ¯ − log[ f T ( Y 1 (1) − Y 0 (1))] . F T ( Y 1 (1) − Y 0 (1))] Y 1 Footprint Node ccdf Node Real Time Matching: All Footprints Not Yet Available Add virtual node: event that transaction is still resident at S 0 Maximum Likelihood ≡ Minimum Weight Perfect Match Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 8 / 20

  42. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  43. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  44. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 Y 1 Footprint Node ccdf Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  45. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 Y 1 Footprint Node ccdf Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  46. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 Y 1 Footprint Node ccdf Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  47. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 No. of unique matches =2 Prob. of correct ML match = 1 2 Y 1 Footprint Node ccdf Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  48. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 No. of unique matches =2 Prob. of correct ML match = 1 2 Y 1 Footprint Node ccdf Node Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  49. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 No. of unique matches =2 Prob. of correct ML match = 1 2 Y 1 Footprint Node ccdf Node Partial Batch Exponential = Worst Case Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  50. Worst Case Maximum Likelihood Performance Definition Minimum probability of correct match over transition time pdfs 1 Worst Case ML Performance= No. of Unique Matches Transaction 1 Transaction 2 Y 0 No. of unique matches =2 Prob. of correct ML match = 1 2 Y 1 Footprint Node ccdf Node Partial Batch Complete Batch Exponential = Worst Case Uniform, Exponential = Worst Case Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 9 / 20

  51. Comparison of Maximum Likelihood with FIFO Time S 0 S 1 FIFO = Match footprint with earliest unmatched transaction Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 10 / 20

  52. Comparison of Maximum Likelihood with FIFO Time S 0 S 1 FIFO = Match footprint with earliest unmatched transaction Always valid, Distribution free, Simpler rule than maximum likelihood Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 10 / 20

  53. Comparison of Maximum Likelihood with FIFO Time S 0 S 1 FIFO = Match footprint with earliest unmatched transaction Always valid, Distribution free, Simpler rule than maximum likelihood Can FIFO and maximum likelihood coincide? Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 10 / 20

  54. Comparison of Maximum Likelihood with FIFO Time S 0 S 1 FIFO = Match footprint with earliest unmatched transaction Always valid, Distribution free, Simpler rule than maximum likelihood Can FIFO and maximum likelihood coincide? Yes: Quadrangle Inequality 0.6 0.4 f T ( t 1 ) f T ( t 2 ) ≥ f T ( t 1 − τ ) f T ( t 2 − τ ) 0.2 0 0 1 2 3 4 5 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 10 / 20

  55. Outline Introduction 1 Two-state System 2 Multi-state System 3 Conclusion 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 11 / 20

  56. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  57. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 IID transitions of transactions S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  58. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 IID transitions of transactions S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Sequence of states visited is a Markov chain 1 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  59. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 IID transitions of transactions S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Sequence of states visited is a Markov chain 1 Transition time depends only on states involved in transition 2 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  60. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 IID transitions of transactions S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Sequence of states visited is a Markov chain 1 Transition time depends only on states involved in transition 2 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  61. Recap of Transaction Model Transaction Model ? ? ? ? ? S 2 S 3 Acyclic semi-Markov process S 1 1 ? 2 IID transitions of transactions S 0 � � S 4 ? 3 � � � � Definition of Semi-Markov Process Sequence of states visited is a Markov chain 1 Transition time depends only on states involved in transition 2 Maximum Likelihood Rule Maximize probability that each footprint is matched correctly to the unique transaction that generated it 1 , . . . , Y π Ns π ML π ML P ( Y π 1 [ˆ 1 , . . . , ˆ N s ] := arg max | Y 0 ) . N s π 1 ,..., π Ns Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  62. Recap of Transaction Model Transaction Model π 2 π 3 π 1 S 2 S 3 Acyclic semi-Markov process S 1 1 2 IID transitions of transactions S 0 S 4 π 4 3 � � � � Definition of Semi-Markov Process Sequence of states visited is a Markov chain 1 Transition time depends only on states involved in transition 2 Maximum Likelihood Rule Maximize probability that each footprint is matched correctly to the unique transaction that generated it 1 , . . . , Y π Ns π ML π ML P ( Y π 1 [ˆ 1 , . . . , ˆ N s ] := arg max | Y 0 ) . N s π 1 ,..., π Ns Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 12 / 20

  63. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  64. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  65. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model 1 2 3 4 S 0 S 1 S 2 S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  66. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model 1 2 3 4 S 0 S 1 S 2 CCDF Node S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  67. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model 1 2 3 4 S 0 S 1 S 2 S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  68. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model 1 2 3 4 S 0 S 1 S 2 S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  69. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 S 1 S 2 S 3 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  70. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 4 3 S 1 5 1 S 2 6 S 3 2 7 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  71. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 4 3 S 1 5 1 S 2 6 S 3 2 7 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  72. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 4 3 S 1 5 1 S 2 6 S 3 2 7 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  73. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 4 3 S 1 5 1 S 2 6 S 3 2 7 S 4 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  74. Special Cases: Linear & Tree Models Linear Model With Snapshot of Footprints 4 3 2 1 ? ? ? ? ... S 0 S 1 S 3 S 4 Maximum Likelihood in Linear Model Tree Model 1 2 3 4 S 0 4 3 S 1 5 1 S 2 6 S 3 2 7 S 4 ML ≡ Series of Bipartite Matchings Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 13 / 20

  75. Acyclic Semi-Markov Process Acyclic Semi Markov Process: Two-Stage Systems No common imm. predecessor: States in end stage of different systems Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 14 / 20

  76. Acyclic Semi-Markov Process Acyclic Semi Markov Process: Two-Stage Systems No common imm. predecessor: States in end stage of different systems 3 4 2 1 5 ({2},{3}) 4 2 5 3 Anandkumar, Bisdikian, Agrawal Monitoring Transactions Using Footprints ACM SIGMETRICS ‘08 14 / 20

Recommend

Inter Internet monitoring net monitoring and web tracking and web tracking Engineering &

Inter Internet monitoring net monitoring and web tracking and web tracking Engineering &

CyLab Inter Internet monitoring net monitoring and web tracking and web tracking Engineering & Public Policy Lorrie Faith Cranor September 30, 2014 y & c S a e v c i u r P r i t e y l b L a a s b U o 8-533

440 views • 42 slides
Rebasing MadeSimple TwoBranches spaghetti master Merge spaghetti master Rebase

Rebasing MadeSimple TwoBranches spaghetti master Merge spaghetti master Rebase

Rebasing MadeSimple TwoBranches spaghetti master Merge spaghetti master Rebase spaghetti master Rebase spaghetti master Rebase spaghetti master Checkout spaghetti master Rebase (withFast-forward)

569 views • 34 slides
Program for Heat-map Entropy Evaluation of Eye-tracking Data Seung-Bin Son a , Yejin Lee a ,

Program for Heat-map Entropy Evaluation of Eye-tracking Data Seung-Bin Son a , Yejin Lee a ,

Transactions of the Korean Nuclear Society Virtual Spring Meeting July 9-10, 2020 Program for Heat-map Entropy Evaluation of Eye-tracking Data Seung-Bin Son a , Yejin Lee a , Hyun-Chul Lee a* a Severe Accident Monitoring and Mitigation Research

186 views • 3 slides
THE FALL 2018 NFL PRIMETIME SEASON & THE SUPER BOWL KTG CONTENT STRATEGY SUPER BOWL

THE FALL 2018 NFL PRIMETIME SEASON & THE SUPER BOWL KTG CONTENT STRATEGY SUPER BOWL

THE FALL 2018 NFL PRIMETIME SEASON & THE SUPER BOWL KTG CONTENT STRATEGY SUPER BOWL HALFTIME PERFORMERS Super Bowl PERFORMERS Super Bowl PERFORMERS 2019 Maroon 5 2012 Madonna 2018 Justin Timberlake 2011 Black Eyed Peas 2017

139 views • 3 slides
Monitoring or Tracking for Monitoring or Tracking for Performance and Performance and

Monitoring or Tracking for Monitoring or Tracking for Performance and Performance and

Monitoring or Tracking for Monitoring or Tracking for Performance and Performance and Compliance Compliance Subrecipients, UGLGs, and Subrecipients, UGLGs, and Contractors Contractors Why monitor? Why monitor? Improve subrecipient and

534 views • 19 slides
Sorting networks using a different architecture? Yep. 2. Spaghetti sort ! Lecture 24 November

Sorting networks using a different architecture? Yep. 2. Spaghetti sort ! Lecture 24 November

NEW CS 473: Theory II, Fall 2015 Model of Computation 1. Q: Perform a computational task considerably faster by Sorting networks using a different architecture? Yep. 2. Spaghetti sort ! Lecture 24 November 19, 2015 1/52 2/52 Spaghetti

761 views • 9 slides
Jefferson County Vision Zero Safety Bowl Mike Vaughn Kentuckys 2019 Safety Bowl

Jefferson County Vision Zero Safety Bowl Mike Vaughn Kentuckys 2019 Safety Bowl

Jefferson County Vision Zero Safety Bowl Mike Vaughn Kentuckys 2019 Safety Bowl National Champion Mini Safety Bowl How Many Highway Fatalities did KY Experience in 2018? A. 724 B. 782 C. 834 Mini Safety Bowl How Many

790 views • 59 slides
ATOM: Automated Tracking, Orchestration and Monitoring of Resource Usage in Infrastructure as a

ATOM: Automated Tracking, Orchestration and Monitoring of Resource Usage in Infrastructure as a

Motivation Design Evaluation Discussion ATOM: Automated Tracking, Orchestration and Monitoring of Resource Usage in Infrastructure as a Service Systems Min Du, Feifei Li School of Computing, University of Utah ATOM: Automated Tracking,

1.28k views • 98 slides
The Highgate Bowl: past, present and future Presentation for Highgate U3A by Maggy

The Highgate Bowl: past, present and future Presentation for Highgate U3A by Maggy

The Highgate Bowl: past, present and future Presentation for Highgate U3A by Maggy Meade-King Trustee of Friends of the Highgate Bowl October 2018 The Bowl in context: Highgates history Hamlet first documented in 14 th century

252 views • 21 slides
Compliance Tracking T ool Activity Entry and Progress Monitoring November 2012 1 Login

Compliance Tracking T ool Activity Entry and Progress Monitoring November 2012 1 Login

Compliance Tracking T ool Activity Entry and Progress Monitoring November 2012 1 Login Add an Activity Progress Monitoring Verify Correction of Noncompliance 2 Compliance Tracking T ool Address

747 views • 36 slides
FY21 21 st Century Community Learning Centers Tracking and Monitoring System Training November

FY21 21 st Century Community Learning Centers Tracking and Monitoring System Training November

FY21 21 st Century Community Learning Centers Tracking and Monitoring System Training November 2020 Welcome and Overview Charmaine Davis-Bey, M.Ed., EdPsyc 21 st CCLC Education Program Specialist Compliance/Monitoring System Lead Regional

908 views • 63 slides
FINANCIAL YEAR ENDED 30 SEPTEMBER 2019 Hollywood Bowl Group plc WELCOME Hollywood Bowl Group plc

FINANCIAL YEAR ENDED 30 SEPTEMBER 2019 Hollywood Bowl Group plc WELCOME Hollywood Bowl Group plc

FINANCIAL YEAR ENDED 30 SEPTEMBER 2019 Hollywood Bowl Group plc WELCOME Hollywood Bowl Group plc AGEN DA STEPHEN BURNS LAURENCE KEEN CEO CFO FY2019 highlights Financial performance Growth strategy IFRS 16 Summary

902 views • 50 slides
Horse Bowl, Horse Judging, and Hippology Developed by: Cecil Co. 4-H Volunteer, Monica Mason

Horse Bowl, Horse Judging, and Hippology Developed by: Cecil Co. 4-H Volunteer, Monica Mason

Horse Bowl, Horse Judging, and Hippology Developed by: Cecil Co. 4-H Volunteer, Monica Mason Horse Bowl In horse bowl, teams of four face off in front of a panel of judges and a moderator. In order to get the chance to answer, competitors must

714 views • 14 slides
Fuel Monitoring Presentation Fuel Monitoring We specialize in fuel monitoring also can customize

Fuel Monitoring Presentation Fuel Monitoring We specialize in fuel monitoring also can customize

Fuel Monitoring Presentation Fuel Monitoring We specialize in fuel monitoring also can customize any tracking product: We provides fuel monitoring comprehensive solutions for Vehicles and Generators with very comprehensive reports: Offering

256 views • 8 slides
Dining Philosophers Five philosophers sit around a table with five forks and spaghetti to eat.

Dining Philosophers Five philosophers sit around a table with five forks and spaghetti to eat.

Dining Philosophers Five philosophers sit around a table with five forks and spaghetti to eat. Philosophers think for a while and they want to eat, only spaghetti, for a while. To eat a philosopher requires two forks, one from the left and one

626 views • 18 slides
GPU POWERED SOLUTIONS IN THE SECOND KAGGLE DATA SCIENCE BOWL SECOND ANNUAL DATA SCIENCE BOWL

GPU POWERED SOLUTIONS IN THE SECOND KAGGLE DATA SCIENCE BOWL SECOND ANNUAL DATA SCIENCE BOWL

April 4-7, 2016 | Silicon Valley GPU POWERED SOLUTIONS IN THE SECOND KAGGLE DATA SCIENCE BOWL SECOND ANNUAL DATA SCIENCE BOWL Massive online data science contest Mon 14 Dec 2015 Mon 14 Mar 2016 192 teams, 293 data scientists finished

805 views • 53 slides
DRSI Sharepoint Monitoring and Audit Folders FOLDERS SAMPLE ATTACHMENTS: Monitoring/TA

DRSI Sharepoint Monitoring and Audit Folders FOLDERS SAMPLE ATTACHMENTS: Monitoring/TA

DRSI Sharepoint Monitoring and Audit Folders FOLDERS SAMPLE ATTACHMENTS: Monitoring/TA Correspondence & Monitoring Reports Event Profiles Tracking of Individual Monitoring Exhibits Related to Individual Findings/Concerns

86 views • 5 slides
TaintDroid: An Information-Flow Tracking System for Realtime Privacy Monitoring on Smartphones

TaintDroid: An Information-Flow Tracking System for Realtime Privacy Monitoring on Smartphones

TaintDroid: An Information-Flow Tracking System for Realtime Privacy Monitoring on Smartphones William Enck Peter Gilbert Byung-Gon Chun Landon P. Cox Jaeyeon Jung Patrick McDaniel Anmol N. Sheth Presentation by Krzysztof Pawlowski

447 views • 30 slides
Nested Transactions Nested Transactions Flat transactions The rules for committing of

Nested Transactions Nested Transactions Flat transactions The rules for committing of

Nested Transactions Nested Transactions Flat transactions The rules for committing of nested transactions Nested Transactions A transaction commits or aborts only after its child transactions have completed; Structured in

808 views • 11 slides
Monitoring Non-Communicable Disease (NCD) Targets and Tracking the Tobacco Epidemic Dr Lubna

Monitoring Non-Communicable Disease (NCD) Targets and Tracking the Tobacco Epidemic Dr Lubna

The Global Burden of Tobacco: Monitoring Non-Communicable Disease (NCD) Targets and Tracking the Tobacco Epidemic Dr Lubna Bhatti Prevention of Noncommunicable Diseases WHO Geneva SESRIC WORKSHOP, Ankara, August 2017 2000: WHO Global

531 views • 35 slides
SMA Teaching Moment Take-A-Stand: Super Bowl Commercial Nov.5, 2016 Jin-Woo Kim, Ph.D.

SMA Teaching Moment Take-A-Stand: Super Bowl Commercial Nov.5, 2016 Jin-Woo Kim, Ph.D.

SMA Teaching Moment Take-A-Stand: Super Bowl Commercial Nov.5, 2016 Jin-Woo Kim, Ph.D. Department of MarkeJng College of Business 2 StaJsJcs: Viewership In 2016, adver.sements in the Super Bowl were watched by 111.9 million U.S. viewers and

464 views • 14 slides
Super Bowl LII Committee of the Whole Briefing December 13, 2017 1 Stakeholders NFL NFL

Super Bowl LII Committee of the Whole Briefing December 13, 2017 1 Stakeholders NFL NFL

CITY OF MINNEAPOLIS Super Bowl LII Committee of the Whole Briefing December 13, 2017 1 Stakeholders NFL NFL vendors Super Bowl Host Committee Partners Third party event organizers 2 The Citys Role Provide

358 views • 10 slides
Project planning and status (May 2018) Ongoing: Multi-camera r/o, control, monitoring and

Project planning and status (May 2018) Ongoing: Multi-camera r/o, control, monitoring and

PANDA pellet tracking Project planning and status (May 2018) Ongoing: Multi-camera r/o, control, monitoring and analysis developments. Preparation of a generator tracking chamber that can house four measurement levels. All hardware

405 views • 5 slides
Tracking System for Monitoring Small Wildlife Species (Drone Project) Presented by: Jason

Tracking System for Monitoring Small Wildlife Species (Drone Project) Presented by: Jason

Unmanned Aerial Radio Tracking System for Monitoring Small Wildlife Species (Drone Project) Presented by: Jason Vizcaino Kellan Rothfus Lauren Adoram-Kershner Lance Eberle Project Description Offer an efficient way to assist radio

735 views • 9 slides

More recommend