Deck 10: Distributed DBMSs-Advanced Concepts and Object-Oriented DBMSs-Concepts and Design

Question 1

The centralized two-phase commit protocol uses a series of timeouts to prevent unnecessary blocking. Discuss the actions for both coordinator and participants when a timeout occurs. Consideration should be given to the various stages of the commit protocol.

Answer 1

See Section 25.4.3.

Question 2

Under the three-phase commit protocol, discuss how the coordinator and participants would recover following a failure. Consideration should be given to the various stages of the commit protocol.

Answer 2

Coordinator Failure
1. Failure in INITIAL state. The coordinator has not yet started the commit procedure. Recovery in this case starts the commit procedure.
2. Failure in WAITING state. The coordinator has sent the prepare message and although has not received all responses, the coordinator has not received an abort response. In this case, recovery restarts the commit procedure.
3. Failure in DECIDED state. The coordinator has instructed the participants to globally abort or commit the transaction. On restart, if the coordinator has received all acknowledgements, it can complete successfully. Otherwise, it has to initiate the termination protocol discussed above.
4. Failure in PRE-COMMITTED state. The coordinator has instructed the participants to pre-commit the transaction. On restart, if the coordinator has received all acknowledgements, it can send the global commit message. Otherwise, it can send the pre-commit message again.

Participant Failure
The objective of the recovery protocol for a participant is to ensure that a participant process on restart performs the same action as all other participants and that this restarting can be done independently (i.e. without the need to consult either the coordinator or the other participants).
1. Failure in INITIAL state. The participant has not yet voted on the transaction. Therefore, on recovery, it can unilaterally abort the transaction, as it would have been impossible for the coordinator to have reached a global commit decision without this participants vote.
2. Failure in PREPARED state. The participant has sent its vote to the coordinator. In this case, recovery is via the termination protocol discussed above.
3. Failure in ABORTED/COMMITTED states. Participant has completed the transaction. Therefore, on restart, no further action is necessary.
4. Failure in PRE-COMMITTED state. Participant informs coordinator and waits for global commit message.

Question 3

Consider six transactions T₁, T₂, T₃, T₄, and T₅ with:
- T₁ initiated at site S₁ and spawning an agent at site S₂,
- T₂ initiated at site S₃ and spawning an agent at site S₁,
- T₃ initiated at site S₁ and spawning an agent at site S₃,
- T₄ initiated at site S₂ and spawning an agent at site S₃,
- T₅ initiated at site S₃.
The locking information for these transactions is shown in Table 1. Produce the local wait-for-graphs (WFGs) for each of the sites. What can you conclude from the local WFGs?

Answer 3

Conclusion: There is no local deadlock at any site.

Question 4

One of the most well-known methods for distributed deadlock detection was developed by Obermarck. Explain how Obermarck's method works and how deadlock is detected and resolved.

Question 5

Using the above transactions, demonstrate how Obermarck's method for distributed deadlock detection works.

Question 6

Discuss why traditional transaction management protocols are too restrictive for advanced database applications.

Question 7

Case Study - Library System
The librarian will access the system to issue books, record reservations, generate return requests for reserved books and record books being returned. Returning a book may involve generating an availability notification if the returned book has been reserved by another member. If an availability notification is generated then the librarian sends it out to the member. When a book is requested to be issued or requested to be reserved then the system validates details on the membership card as well as validating that the requested book is one stocked by the library. If the membership card is invalid or the library does not stock the book then the request is rejected. If a book is requested for issue but is on-loan then that book will be reserved. Members of the library can display details of books stocked. The subscriptions section deals with membership issues e.g. renewing membership cards on an annual basis etc. The purchasing section deals with adding new books to the library.
There may be multiple copies of books held in the library. When a book is borrowed the borrowing date is noted and when the book is returned the return date is noted. The details held about each reservation include who the current borrower(s) is and who has reserved the book.
When a copy of a book arrives from the suppliers, it is held in storage until its details are registered on the system. These details include the number, author and title of the book. Once registered the copy is put in the lending shelf which means that it is available to be borrowed. When a copy is returned, if the book has been reserved then the copy is held in a reservation area otherwise it is returned to the lending shelf. If a copy is reported lost by the borrower then the copy details are deleted from the system by the librarian.
A reservation
Reservation
Number:
Date:
Book
Number:
Title:
Author:
Reserver
Number:
Name:
Address:
Current Borrower
Number:
Name:
Address:

-Produce use case diagrams and a set of associated sequence diagrams for the above case study.

Question 8

Give your definition of an Object-Oriented Database Management System (OODBMS). Discuss what you consider to be the three most important advantages of a DDBMS over a relational DBMS. Justify your selection of the three advantages.

Question 9

Despite the superior expressive power of the Object-Oriented Database Management System (OODBMS) in comparison to the established relational systems, the acceptance of the OODBMS will ultimately depend on its performance. The key to this may well lie with how persistent objects are accessed. Discuss the design goals for the incorporation of persistence in a programming language.

Question 10

Object databases have roots in both programming languages and database management. However, not all aspects of these two technologies blend together easily. One area of potential conflict arises when we try to completely separate persistence and type.
(a) Define the orthogonality of persistence and type.
(b) Discuss this issue for any three of the following subsystems:
(1) Queries
(2) Schemas
(3) Transactions
(4) Existence Semantics.

Question 11

Discuss the concept of object identifiers (OIDs) in an object DBMS and discuss four different approaches for the representation of OIDs.

Question 12

Discuss how OIDs differ from C++ pointers.

Question 13

(a) 'Pointer swizzling' is a technique that can be used to optimize access to main memory resident persistent objects. Discuss in general terms how pointer swizzling works.(b) Classify and evaluate the different approaches to pointer swizzling.(c) Briefly discuss any alternative approach that could be used to handle persistence in an ODBMS.

Question 14

(a) Discuss why traditional transaction management protocols are too restrictive for advanced database applications.(b) 'Only Object-Oriented Database Management Systems can support alternative transaction management protocols effectively.' Discuss the validity of this statement.