1 / 43

SAMC: Sematic-Aware Model Checking for Fast Discovery of Deep Bugs in Cloud Systems

SAMC: Sematic-Aware Model Checking for Fast Discovery of Deep Bugs in Cloud Systems. Tanakorn Leesatapornwongsa , Mingzhe Hao , Pallavi Joshi * , Jeffrey F. Lukman † , and Haryadi S. Gunawi. †. *. Internet Services. Internet Services. Cloud Systems. Reliability. Complex failures.

malory
Download Presentation

SAMC: Sematic-Aware Model Checking for Fast Discovery of Deep Bugs in Cloud Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. SAMC: Sematic-Aware Model Checkingfor Fast Discovery ofDeep Bugs in Cloud Systems Tanakorn Leesatapornwongsa,MingzheHao, Pallavi Joshi*,Jeffrey F. Lukman†, and Haryadi S. Gunawi † *

  2. Internet Services

  3. Internet Services Cloud Systems

  4. Reliability Complex failures “Deep bugs”

  5. Deep Bug Example B C A ZooKeeper (synchronization service) Issue #335. 1. Nodes A, B, C start (w/ latex txid: 10) 2. B becomes leader 3. B crashes 4. C becomes leader 5. C commits new txid-value pair (11, X) 6. Acrashes, beforecommitting the new txid 11 7. C loses quorum and C crashes 8. A and B are back online afterC crashes 9. Abecomes leader 10. A's commits new txid-value pair (11, Y) 11. C is back online afterA's new tx commit 12. C announce to B (11, X) 13. B replies diff starting with tx 12 14. Inconsistency: A has (11, Y), C has (11, X) F F F F F F F F L L L L L L L x x x x y y x x y y PERMANENT INCONSISTENT REPLICA

  6. Deep Bug Characteristics ZooKeeper (synchronization service) Issue #335. 1. Nodes A, B, C start (w/ latex txid: 10) 2. B becomes leader 3. B crashes 4. C becomes leader 5. C commits new txid-value pair (11, X) 6. Acrashes, beforecommitting the new txid 11 7. C loses quorum and C crashes 8. A and B are back online afterC crashes 9. Abecomes leader 10. A's commits new txid-value pair (11, Y) 11. C is back online afterA's new tx commit 12. C announce to B (11, X) 13. B replies diff starting with tx 12 14. Inconsistency: A has (11, Y), C has (11, X) 1. Out-of-order messages 2. Multiple crashes Specific Order 3. Multiple reboots HAPPEN IN ANY ORDER 1 2 3

  7. Study of Deep Bugs # crashes / # reboots # crashes / # reboots SEVERE IMPLICATIONS INCONSISTENT REPLICAS, DATA LOSS, DOWNTIMES, ETC. x-axis is bug number y-axis is number of crashes and reboots # crashes / # reboots

  8. How do we catch deep bugs indistributed systems?

  9. How to Catch Deep Bugs • Distributed system model checker • Re-ordering all non-deterministic events • Find which specific orderings lead to bugs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 2 7 1 4 5 6 3 8 11 10 9 12 14 13 6 9 3 4 5 1 7 8 2 10 11 13 12 14 3 4 1 5 2 6 7 8 9 12 11 10 14 13 2 1 3 4 5 7 6 8 9 10 11 14 12 13 . ..

  10. What’s Wrong withExisting Model Checkers? • Last 7 years • MaceMC[NSDI’07], Modist [NSDI’09], dBug [SSV’10], Demeter [SOSP’13], etc. • BUT • Too many events • Multiple crashes and reboots • Create more messages • No model checker incorporate multiple crashes and reboots • Cannot find deep bugs! 100 events

  11. How do we catch deep bugsREALLY FAST?

  12. Black-Box Approach • Existing model checkers are so slow • They treat target systems as black boxes • A large number of event orderings are generated Black Box A C D B

  13. Semantic Knowledge • How can we make model checker fast? • Exploit semantic knowledge • Semantic-aware model checker (SAMC) Black Box SAMC A C D B

  14. Black Box vs. SAMC Black Box Message Processing Semantic A C D B A C D B Unnecessary Re-orderings Lead to the same state

  15. SAMC with Crashes Crash Recovery Semantic N2 A,B N1 N3 C,D Unnecessary Re-orderings N4

  16. Generic Reduction Algorithms Reboot Synchronization Symmetry (RSS) Crash-Message Independence (CMI) Local-Message Independence (LMI) CrashRecovery Symmetry (CRS) Principle of Semantic Awareness

  17. SAMC Implementation and Integration • SAMC implementation • 10,000 LOC from scratch • Apply SAMC to 3 cloud systems • 7 protocols • 10 versions

  18. Result • Reproduced 12 old bugs • Compare to state-of-the-art techniques • Dynamic Partial Order Reduction (DPOR) • Random-DPOR • Find bugs 2xto 340xfaster • 49x on average • Found 2 new bugs • Submit them to developers

  19. Outline • Intuition • SAMC • Local-Message Independence • Crash-Message Independence • Crash Recovery Symmetry • Reboot Synchronization Symmetry • Evaluation

  20. Dependency vs. Independency A, B Node State = S S’ A, B S S’ S S’’ B, A B, A A B Unnecessary A, B = Dependent A, B = Independent INDEPENDENT = NO REORDERING 2X SPEED-UP

  21. Black Box vs. SAMC Black Box Semantic Info A B C D A B C D All dependent Dependent Dependent 6X SPEED-UP

  22. Reduction Speed-up ABCD ABCD ABDC ABDC S S ACBD ACBD . . . . . . . . . . . . . . . . . . . . . . . .

  23. How to Declare Message Independency? Q: Which concurrent messages are independent ? A: Use message processing semantic

  24. Message Processing Semantic inSimplified Leader Election B = 3 B = 3 B = 3 B = 3 B = 3 B = 4 Belief = n3 V = 1 V = 2 V = 4 if(vote <= belief) // do nothing else belief = vote; Vote=1 Vote=2 Vote=4

  25. Removing Re-ordering viaMessage Processing Semantic 1, 2, 3 1, 3, 2 2, 1, 3 • B = 4 B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 • B = 4 Belief=4 V = 3 V = 3 • V = 2 • V = 2 V = 1 • V = 1 V = 3 V = 2 • V = 1 Vote=2 Vote=3 Vote=1 . . . Unnecessary if (vote <= belief) // do nothing else belief = vote;

  26. Formalizing the Intuition MESSAGE PROCESSING SEMANTIC if (vote <= belief) // do nothing else belief = vote; DISCARD PATTERN if (isDiscard(msg, ls)) { // do nothing; } DISCARD PREDICATE booleandiscardPredicate(msg, ls) { if (msg.vote <= ls.belief) return true; else return false; }

  27. Formalizing the Intuition Belief=4 Vote=2 Vote=3 Vote=1

  28. Outline • Intuition • SAMC • Local-Message Independence • Crash-Message Independence • Crash Recovery Symmetry • Reboot Synchronization Symmetry • Evaluation

  29. SAMC Architecture a, b SAMC Local state: ls1 Protocol Specific Rules … … Leader Election Atomic Broadcast interceptor Generic Reduction Policies release(c) RSS LMI CMI CRS interceptor Basic Reduction Techniques Dynamic PartialOrder Reduction (DPOR) Symmetry Local state: ls2 c, d

  30. Local-Message Independence SAMC Generic ReductionPolicies Local-Message Independence (LMI) Crash-Message Independence (CMI) Reboot Synchronization Symmetry (RSS) Crash Recovery Symmetry (CRS)

  31. Local-Message Independence • Discard pattern • Increment pattern • Constant pattern C=0 C=1 C=2 C=0 C=2 C=1 if (msg.type== ack) { node.ackCount++; } ack ack ack ack booleanincrementPredicate(msg, ls) { if (msg.type== ack) return true; else return false; }

  32. Crash-Message Independence SAMC Generic ReductionPolicies Local-Message Independence (LMI) Crash-Message Independence (CMI) Reboot Synchronization Symmetry (RSS) Crash Recovery Symmetry (CRS)

  33. Crash-Message Independence L void handleCrash() { if (X == follower && isQuorum()) followerCount--; // No new messages } A,B C,D F F F booleanlocalImpact(X, ls) { if (X == follower && isQuorum()) return true; else return false; } L Local Impact A,B C,D F F F

  34. Crash-Message Independence L void handleCrash() { if (X == leader || !isQuorum()) electLeader() // New messages created } A,B C,D F F F booleanglobalImpact(X, ls) { if (X == leader || !isQuorum()) return true; else return false; } L Global Impact S S S

  35. Crash Recovery SymmetryReboot Synchronization Symmetry SAMC Generic ReductionPolicies Crash-Message Independence (CMI) Local-Message Independence (LMI) Crash Recovery Symmetry (CRS) Reboot Synchronization Symmetry (RSS)

  36. Outline • Intuition • SAMC • Local-Message Independence • Crash-Message Independence • Crash Recovery Symmetry • Reboot Synchronization Symmetry • Evaluation

  37. Evaluation • SAMC implementation • 10,000 LOC from scratch • Apply SAMC to 3 cloud systems • 7 protocols • 10 versions

  38. Protocol-Specific Rules(e.g. ZooKeeper Leader Election) • Guide SAMC to remove re-orderings • 35 LOC on average per protocol

  39. Catching Old Bugs A table shows number of executions to reach the bugs and speedup

  40. Catching Old Bugs A table shows number of executions to reach the bugs and speedup

  41. Reduction Ratio • ZooKeeperleader election protocol • Run black-box DPOR • After each execution, find that this execution is executed by SAMC or not • Count DPOR’s executions that are executed by SAMC too

  42. Conclusion • Deep bugs live in the cloud • Model checker needs to incorporate complex failure to reach deep bugs • State space explosion • Semantic-aware model checking • LMI, CMI, CRS, RSS • Bring future research questions • What other semantic knowledge is useful? • How to extract them from the code automatically?

  43. Thank YouQuestions? http://ucare.cs.uchicago.edu

More Related