1. 15.1

2. 15.2 General Overview Where we’ve been... DBA skills for relational DB’s: Logical Schema Design  E/R diagrams  Decomposition and Normalization Query languages  RA, RC, SQL Integrity Constraints Transactions Database Implementation Issues Files, buffering, indexing Today: Data Integrity

3. 15.3 What Does a DBMS Manage? What Does a DBMS Manage? 1. Data organization  E/R Model  Relational Model  2. Data Retrieval  Relational Algebra  Relational Calculus  SQL  3. Data Integrity  Integrity Constraints  Transactions

4. 15.4

5. 15.5 Updates in SQL An example: UPDATE account SET balance = balance -50 WHERE acct_no = A102 account Dntn: A102: 300 Dntn: A15: 500 Mian: A142: 300 … … (1) Read (2) update (3) write Transaction: 1.Read(A) 2.A <- A -50 3.Write(A) What takes place: memory Disk

6. 15.6 The Threat to Data Integrity Consistent DB Name Acct bal -------- ------ ------ Joe A-33 300 Joe A-509 100 Joe’s total: 400 Consistent DB Name Acct bal -------- ------ ------ Joe A-33 250 Joe A-509 150 Joe’s total: 400 transaction Inconsistent DB Name Acct bal -------- ------ ------ Joe A-33 250 Joe A-509 100 Joe’s total: 350 What a Xaction should look like to Joe What actually happens during execution

7. 15.7 Transactions What?:  Updates to db that can be executed concurrently Why?: (1) Updates can require multiple reads, writes on a db e.g., transfer $50 from A-33 to A509 = read(A) A  A -50 write(A) read(B) B  B+50 write(B) (2) For performance reasons, db’s permit updates to be executed concurrently Concern: concurrent access/updates of data can compromise data integrity

8. 15.8 ACID Properties Atomicity: either all operations in a Xaction take effect, or none Consistency: operations, taken together preserve db consistency Isolation: intermediate, inconsistent states must be concealed from other Xactions Durability. If a Xaction successfully completes (“commits”), changes made to db must persist, even if system crashes Properties that a Xaction needs to have:

9. 15.9 Demonstrating ACID Transaction to transfer $50 from account A to account B: 1.read(A) 2.A := A – 50 3.write(A) 4.read(B) 5.B := B + 50 6.write(B) Consistency: total value A+B, unchanged by Xaction Atomicity: if Xaction fails after 3 and before 6, 3 should not affect db Durability: once user notified of Xaction commit, updates to A,B should not be undone by system failure Isolation: other Xactions should not be able to see A, B between steps 3-6

10. 15.10 Threats to ACID 1. Programmer Error e.g.: $50 substracted from A, $30 added to B  threatens consistency 2. System Failures e.g.: crash after write(A) and before write(B)  threatens atomicity e.g.: crash after write(B)  threatens durability 3. Concurrency E.g.: concurrent Xaction reads A, B between steps 3-6  threatens isolation

11. 15.11 Isolation Simplest way to guarantee: forbid concurrent Xactions! But, concurrency is desirable: (1) Achieves better throughput (TPS: transactions per second) one Xaction can use CPU while another is waiting for disk to service request (2) Achieves better average response time short Xactions don’t need to get stuck behind long ones Prohibiting concurrency is not an option

12. 15.12 Isolation Approach to ensuring Isolation:  Distinguish between “good” and “bad” concurrency  Prevent all “bad” (and sometime some “good”) concurrency from happening OR  Recognize “bad” concurrency when it happens and undo its effects (abort some transactions)  Pessimistic vs Optimistic CC Both pessimistic and optimistic approaches require distinguishing between good and bad concurrency How: concurrency characterized in terms of possible Xaction “schedules”

13. 15.13 Schedules Schedules – sequences that indicate the chronological order in which instructions of concurrent transactions are executed  a schedule for a set of transactions must consist of all instructions of those transactions  must preserve the order in which the instructions appear in each individual transaction T1 1 2 3 T2 A B C D T1 1 2 3 T2 A B C D one possible schedule:

14. 15.14 Example Schedules Constraint: The sum of A+B must be the same Before: 100+50 After: 45+105 T1 read(A) A <- A -50 write(A) read(B) B<-B+50 write(B) T2 read(A) tmp <- A*0.1 A <- A – tmp write(A) read(B) B <- B+ tmp write(B) Transactions: T1: transfers $50 from A to B T2: transfers 10% of A to B =150, consistent Example 1: a “serial” schedule

15. 15.15 Example Schedule Another “serial” schedule: T1 read(A) A <- A -50 write(A) read(B) B<-B+50 write(B) T2 read(A) tmp <- A*0.1 A <- A – tmp write(A) read(B) B <- B+ tmp write(B) Before: 100+50 After: 40+110 Consistent but not the same as previous schedule.. Either is OK! =150, consistent

16. 15.16 Example Schedule (Cont.) Another “good” schedule: T1 read(A) A <- A -50 write(A) read(B) B<-B+50 write(B) T2 read(A) tmp <- A*0.1 A <- A – tmp write(A) read(B) B <- B+ tmp write(B) Effect: Before After A 100 45 B 50 105 Same as one of the serial schedules Serializable

17. 15.17 Example Schedules (Cont.) A “bad” schedule Before: 100+50 = 150 After: 50+60 = 110 !! Not consistent T1 read(A) A <- A -50 write(A) read(B) B<-B+50 write(B) T2 read(A) tmp <- A*0.1 A <- A – tmp write(A) read(B) B <- B+ tmp write(B)

18. 15.18 Example Schedule (Schedule B) T 1 T 2 T 3 T 4 T 5 read(X) read(Y) read(Z) read(V) read(W) read(Y) write(Y) write(Z) read(U) read(Y) write(Y) read(Z) write(Z) read(U) write(U)

19. 15.19 Transaction Definition in SQL Data manipulation language must include a construct for specifying the set of actions that comprise a transaction. In SQL, a transaction begins implicitly. A transaction in SQL ends by:  Commit work commits current transaction and begins a new one.  Rollback work causes current transaction to abort. Levels of consistency specified by SQL-92:  Serializable — default (more conservative than conflict serializable)  Repeatable read  Read committed  Read uncommitted

20. 15.20 Transactions in SQL Serializable — default - can read only committed records - if T is reading or writing X, no other Xaction can change X until T commits - if T is updating a set of records (identified by WHERE clause), no other Xaction can change this set until T commits Weaker versions (non-serializable) levels can also be declared Idea: tradeoff: More concurrency => more overhead to ensure valid schedule. Lower degrees of consistency useful for gathering approximate information about the database, e.g., statistics for query optimizer.

21. 15.21

22. 15.22 General Overview Relational model - SQL  Formal & commercial query languages Functional Dependencies Normalization Physical Design Indexing Query Processing and Optimization Transaction Processing and CC

23. 15.23 Transaction Concept A transaction is a unit of program execution that accesses and possibly updates various data items. A transaction must see a consistent database. During transaction execution the database may be inconsistent. A transaction ends with a Commit or an Abort When the transaction is committed, the database must be consistent. State 1State 2

24. 15.24 Prevent P(S) cycles from occurring using a concurrency control manager: ensures interleaving of operations amongst concurrent xactions only result in serializable schedules. T 1 T 2 ….. T n CC Scheduler How to enforce serializable schedules? Serializable schedule

25. 15.25 Concurrency Via Locks Idea:  Data items modified by one xaction at a time Locks  Control access to a resource  Can block a xaction until lock granted  Two modes:  Shared (read only)  eXclusive (read & write)

26. 15.26 Granting Locks Requesting locks  Must request before accessing a data item Granting Locks  No lock on data item? Grant  Existing lock on data item?  Check compatibility:  Compatible? Grant  Not? Block xaction sharedexclusive sharedYesNo exclusiveNo

27. 15.27 Lock instructions New instructions - lock-S: shared lock request - lock-X: exclusive lock request - unlock: release previously held lock Example: lock-X(B) read(B) B  B-50 write(B) unlock(B) lock-X(A) read(A) A  A + 50 write(A) unlock(A) lock-S(A) read(A) unlock(A) lock-S(B) read(B) unlock(B) display(A+B) T1 T2

28. 15.28 Locking Issues Starvation  T1 holds shared lock on Q  T2 requests exclusive lock on Q: blocks  T3, T4,..., Tn request shared locks: granted  T2 is starved! Solution? Do not grant locks if other xaction is waiting

29. 15.29 Locking Issues No xaction proceeds: Deadlock - T1 waits for T2 to unlock A - T2 waits for T1 to unlock B T1T2 lock-X(B) read(B) B  B-50 write(B) lock-X(A) lock-S(A) read(A) lock-S(B) More later…

30. 15.30 Locking Issues Does not ensure serializability by itself: lock-X(B) read(B) B  B-50 write(B) unlock(B) lock-X(A) read(A) A  A + 50 write(A) unlock(A) lock-S(A) read(A) unlock(A) lock-S(B) read(B) unlock(B) display(A+B) T1 T2 T2 displays 50 less!!

31. 15.31 Two-Phase Locking: 2PL Each xaction has two phases  Growing : only lock request  Shrinking: only unlocks  xactions start in growing phase  First unlock begins the shrinking phase Ensures serializabile schedules !  Order concurrent xactions on the moment of final lock is granted

32. 15.322PL Example: T1 in 2PL T1 lock-X(B) read(B) B  B - 50 write(B) lock-X(A) read(A) A  A - 50 write(A) unlock(B) unlock(A) Growing phase Shrinking phase

33. 15.33 2PL Issues 2PL does not prevent deadlock > 2 xactions involved? - Rollbacks expensive T1T2 lock-X(B) read(B) B  B-50 write(B) lock-X(A) lock-S(A) read(A) lock-S(B)

34. 15.34 Dealing with Deadlocks How do you detect a deadlock?  Wait-for graph  Directed edge from Ti to Tj  Ti waiting for Tj T1T2T3T4 S(V) X(V) S(W) X(Z) S(V) X(W) T1 T2 T4 T3 Suppose T4 requests lock-S(Z)....

35. 15.35 Detecting Deadlocks Wait-for graph has a cycle  deadlock T2, T3, T4 are deadlocked T1 T2 T4 T3 Build wait-for graph, check for cycle How often? - Tunable Expect many deadlocks or many xactions involved Run often to avoid aborts Else run less often to reduce overhead

36. 15.36 Recovering from Deadlocks Rollback one or more xaction  Which one?  Rollback the cheapest ones  Cheapest ill-defined  Was it almost done?  How much will it have to redo?  Will it cause other rollbacks?  How far?  May only need a partial rollback  Avoid starvation  Ensure same xaction not always chosen to break deadlock

37. 15.37

38. 15.38 Review: The ACID properties A A tomicity: All actions in the Xaction happen, or none happen. C C onsistency: If each Xaction is consistent, and the DB starts consistent, it ends up consistent. I I solation: Execution of one Xaction is isolated from that of other Xacts. D D urability: If a Xaction commits, its effects persist. CC guarantees Isolation and Atomicity. The Recovery Manager guarantees Atomicity & Durability.

39. 15.39 Why is recovery system necessary? Transaction failure :  Logical errors: application errors (e.g. div by 0, segmentation fault)  System errors: deadlocks  Aborts System crash:  hardware/software failure causes the system to crash. Disk failure:  head crash or similar disk failure destroys all or part of disk storage The lost data can be in main memory or in disk

40. 15.40 Storage Media Volatile storage:  does not survive system crashes  examples: main memory, cache memory Nonvolatile storage:  survives system crashes  examples: disk, tape, flash memory, non-volatile (battery backed up) RAM Stable storage:  a “mythical” form of storage that survives all failures  approximated by maintaining multiple copies on distinct nonvolatile media

41. 15.41 Recovery and Durability To achieve Durability: Put data on stable storage To approximate stable storage make two copies of data

42. 15.42 Stable-Storage Implementation Solution:  Write to the first disk  Write to the second disk when the first disk completes  The process is complete only after the second write completes successfully Recovery (from disk failures, etc):  Detect bad blocks with the checksum (e.g. parity)  Two good copies, equal blocks: done  One good, one bad : copy good to bad  Two bad copies: ignore write  Two good, unequal blocks? Ans: Copy the second to the first

43. 15.43 Recovery and Atomicity Example: transfer $50 from account A to account B  goal is either to perform all database modifications made by T i or none at all. Requires several inputs (reads) and outputs (writes) Failure after output to account A and before output to B….  DB is corrupted!

44. 15.44 Recovery Algorithms Recovery algorithms are techniques to ensure database consistency and transaction atomicity and durability despite failures Recovery algorithms have two parts 1.Actions taken during normal transaction processing to ensure enough information exists to recover from failures 2.Actions taken after a failure to recover the database contents to a state that ensures atomicity, consistency and durability ARIES: Algorithms for Recovery and Isolation Exploiting Semantics http://en.wikipedia.org/wiki/Algorithms_for_Recovery_and_Isolation _Exploiting_Semantics

45. 15.45 Log-Based Recovery Simplifying assumptions:  Transactions run serially  logs are written directly on the stable storage Log: a sequence of log records; maintains a record of update activities on the database. (Write Ahead Log, W.A.L.) W.A.L. Write-Ahead Log:  record the operation on the log, before you write it on the db.  Record everything else on the log before commit

46. 15.46 Log based approach Log records for transaction Ti: Two approaches using logs Deferred database modification Immediate database modification

47. 15.47 Log example Transaction T1 Read(A) A =A-50 Write(A) Read(B) B = B+50 Write(B) Log

48. 15.48 Deferred Database Modification T i starts: write a record to log. T i write(X)  write to log: V is the new value for X  The write is deferred  Note: old value is not needed for this scheme T i partially commits:  Write to the log DB updates by reading and executing the log:  ……

49. 15.49 Deferred Database Modification How to use the log for recovery after a crash? Redo: if both and are there in the log. Crashes can occur while  the transaction is executing the original updates, or  while recovery action is being taken example transactions T 0 and T 1 (T 0 executes before T 1 ): T 0 : read (A) T 1 : read (C) A: - A - 50 C:-C- 100 Write (A) write (C) read (B) B:- B + 50 write (B)

50. 15.50 Deferred Database Modification (Cont.) Below we show the log as it appears at three instances of time. (a) (b) (c)

51. 15.51 Immediate Database Modification Database updates of an uncommitted transaction is allowed Tighter logging rules are needed to ensure transactions are undoable  Write records must be of the form:  log record must be written before database item is written  Output of DB blocks can occur:  Before or after commit  In any order

52. 15.52 Immediate Database Modification Example Log Write Output A = 950 B = 2050 C = 600 B B, B C B A Note: B X denotes block containing X.

53. 15.53 Immediate Database Modification (Cont.) Recovery procedure :  Undo : is in the log but is not. Undo:  restore the value of all data items updated by T i to their old values, going backwards from the last log record for T i  Redo: and are both in the log.  Redo: sets the value of all data items updated by T i to the new values, going forward from the first log record for T i Both operations must be idempotent: even if the operation is executed multiple times the effect is the same as if it is executed once

54. 15.54 I M Recovery Example (a) (b) (c) Recovery actions in each case above are: (a) undo (T 0 ): B is restored to 2000 and A to 1000. (b) undo (T 1 ) and redo (T 0 ): C is restored to 700, and then A and B are set to 950 and 2050 respectively. (c) redo (T 0 ) and redo (T 1 ): A and B are set to 950 and 2050 respectively. Then C is set to 600

55. 15.55 Checkpoints Problems in recovery procedure as discussed earlier : 1.searching the entire log is time-consuming 2.we might unnecessarily redo transactions which have already output their updates to the database. How to avoid redundant redoes?  Put marks in the log indicating that at that point DB and log are consistent. Checkpoint!

56. 15.56 Checkpoints At a checkpoint: Output all log records currently residing in main memory onto stable storage. Output all modified buffer blocks to the disk. Write a log record onto stable storage.

57. 15.57 Checkpoints (Cont.) Recovering from log with checkpoints: 1.Scan backwards from end of log to find the most recent record 2.Continue scanning backwards till a record is found. 3.Need only consider the part of log following above start record. Why? 4.After that, recover from log with the rules that we had before.

58. 15.58 Example of Checkpoints TcTc TfTf T1T1 T2T2 T3T3 T4T4 checkpoint system failure checkpoint T 1 can be ignored (updates already output to disk due to checkpoint) T 2 and T 3 redone. T 4 undone

59. 15.59 Recovery With Concurrent Transactions To permit concurrency:  All transactions share a single disk buffer and a single log  Concurrency control: Strict 2PL :i.e. Release eXclusive locks only after commit.  Logging is done as described earlier. The checkpointing technique and actions taken on recovery have to be changed  since several transactions may be active when a checkpoint is performed.