1 / 34

Lock-Free Consistency Control for Web 2.0 Applications

Lock-Free Consistency Control for Web 2.0 Applications. Jiang-Ming Yang, Hai-Xun Wang, Ning Gu, Yi-Ming Liu, Chun-Song Wang, Qi-Wei Zhang 25 April 2008. Outline. Motivation & Challenges Problem Setting Our Solutions Experiments Conclusions. Outline. Motivation & Challenges

Download Presentation

Lock-Free Consistency Control for Web 2.0 Applications

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. Lock-Free Consistency Control for Web 2.0 Applications Jiang-Ming Yang, Hai-Xun Wang, Ning Gu, Yi-Ming Liu, Chun-Song Wang, Qi-Wei Zhang 25 April 2008

  2. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Experiments • Conclusions

  3. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Experiments • Conclusions

  4. Motivation • Single Site

  5. Motivation • Multiple Mirror Site

  6. Challenges • How to merge the conflicts in Multiple Mirror Site for Web 2.0 Application? ? ?

  7. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Experiments • Conclusions

  8. Problem Setting • Data shared in mirrored sites can be in varied forms. Here, we assume the shared data on mirrored sites are XML documents. • Consequently, operations on the data are expressed by XML queries and updates.

  9. Problem Setting <Root> <book @title="Introduction to Algorithm"> <category>CS</category> <tag>Hot</tag> </book> <book @title="Advanced Statistical Learning"> <category>UnKnow</category> </book> <book @title="Linear Algebra"> <category>Math</category> </book> </Root>

  10. Problem Setting • CAUSAL RELATIONSHIPS • TRANSACTIONS • CONCURRENT OPERATIONS (Lock Free)

  11. Problem Setting • CAUSAL RELATIONSHIPS U1 Change the title “Advanced Statistical Learning” to “Statistical Learning”. U2 Set the category of the “Statistical Learning” book to “Math”.

  12. Problem Setting • TRANSACTIONS • Using transaction model to execute some critical operations • Should be executed serialized with others

  13. Problem Setting • CONCURRENT OPERATIONS U3 Add a “Discount” tag to books in “Math” category. U4 Set the category of the “Linear Algebra” book to “CS”.

  14. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Causality Preservation • Transaction • Consistency Control • Experiments • Conclusions

  15. Causality Preservation • Definition. (Causal Ordering Relation “→”). Given two operations Oa and Ob from local replica sites i and j respectively, we have Oa → Ob, if and only if (1) i = j, and Oa is generated before Ob is generated; (2) i ≠ j, and Oa is executed on site j before Obis generated; (3) there exists an operation Ox, such that Oa → Oxand Ox → Ob. • Definition.(Concurrent Relation “ ”). Given two operations Oa and Ob, we say Oa and Ob are concurrent or Oa Ob iff neither Oa → Ob, nor Ob → Oa.

  16. Causality Preservation • Vector based Timestamp • SV = <SV1, SV2, . . . , SVn> • Both for Site & operation • Definition. (Execution Condition).Operation O (from site i) is causally ready for execution at site j (i ≠ j) if the following conditions are satisfied: (in reference [6, 14]) • SVO[i] = SVj[i] + 1 • SVO[k] ≤ SVj[k], for all 1 ≤ k ≤ N and k ≠ i.

  17. Transaction • The transaction model is introduced to achieve concurrent transparency • Definition. (Serialized Transaction). Let T be a transaction, and O be an operation (O may or may not be a transaction). Transaction T is a serialized transaction only if either O is executed before T in all sites, or O is executed after T in all sites.

  18. Transaction • O1 → NOOP1 → O3 and O2 → NOOP2 → O4

  19. Transaction • Definition. (TOrder: total order for transactions). It’s a logical order among transactions : T1 ≺ T2 ≺ T3 ≺ … ≺ Tn.

  20. Consistency Control – Basic Idea

  21. Consistency Control - Operation • XPath : /root/book • XQuery & XUpdate FOR $title in /root//title $category = $book/category WHERE $title = "Advanced Statistical Learning" ..

  22. Consistency Control– Storage Model

  23. Consistency Control– Storage Model

  24. ConsistencyControl–StorageModel

  25. Consistency Control- Query Process • It is clear that there is no overlap in their timestamps, and at any time only one of them is valid. • In general, only nodes whose timestamp (tcreate, tdelete) satisfies t ∈ (tcreate, tdelete) are valid at t.

  26. Consistency Control- Query Process

  27. Consistency Control – Size of OHL • In our approach, we store operations in an Operation History List (OHL). • Each time a replica receives a new remote operation, it will update OHL. • When an operation is executed on all replicas, it will be removed from OHL and its related state information will be removed from the inverted list.

  28. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Experiments • Conclusions

  29. Experiments

  30. Experiments

  31. Experiments

  32. Outline • Motivation & Challenges • Problem Setting • Our Solutions • Experiments • Conclusions

  33. Conclusions • We proposed a lock-free approach for consistency maintance in Web 2.0 environment. • We do not use the locking mechanism, so concurrent operations are executed as soon as possible upon their arrival with ensuring the convergence. • We also support the transaction semantics for critical operations without using the locking mechanism relying on the causality preservation approach. • Better load balance, high-speed access and shorter respond time

  34. Thank You !

More Related