Concurrency Control. Chapter 17. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 1

Transcription

1 Concurrency Control Chapter 17 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 1 Conflict Schedules Two actions conflict if they operate on the same data object and at least one of them is a write. The outcome of a schedule depends only on the order of conflicting operations. Two schedules are conflict equivalent if: Involve the same actions of the same transactions Every pair of conflicting actions is ordered the same way Nonconflicting operations of these two schedules could be swapped. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 2 1

2 Conflict Serializable Schedules Schedule S is conflict serializable if S is conflict equivalent to some serial schedule. Every conflict serializable schedule is serializable. Some serializable schedules are not conflict serializable. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 3 Conflict Serializable Schedules T1 T2 T3 R(A) W(A) W(A) W(A) The schedule is equivalent to serially executing T1, T2, T3. The schedule is not conflict serializable to the serial schedule because T1 and T2 write on object A in different order. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 4 2

3 Precedence Graph All potential conflicts between transactions in a schedule are captured in it. A node for each committed transaction in S. An arc from Ti to Tj if an action of Ti precedes and conflicts with one of Tj s actions. Ti reads X before Tj writes X Ti writes X before Tj reads X Ti writes X before Tj writes X Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 5 Example T1: X(A),R(A), W(A),X(B),R(B), W(B), C T2: X(A),R(A), W(A), X(B),R(B), W(B), C T1 T2 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 6 3

4 T1 S(A) R(A) X(C) R(C) W(C) T2 S(A) R(A) X(B) R(B) W(B) T1 T2 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 7 T1 T2 T3 R(A) W(A) W(A) W(A) T1 T3 T2 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 8 4

5 Acyclic Precedence Graph Theorem: A schedule is conflict serializable if and only if its precedence graph is acyclic. T1: R(A), W(A), R(B), W(B) T2: R(A), W(A), R(B), W(B) T1 A T2 Precedence graph B The cycle in the graph reveals the problem. The output of T1 depends on T2, and viceversa. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 9 Example T1 T2 T3 R(A) R(B) W(A) R(A) W(B) W(A) R(B) W(B) T1 T3 T2 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 10 5

6 Example T1 T2 T3 R(A) R(B) W(A) R(B) R(A) W(B) W(A) W(B) T1 T3 T2 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 11 Review: Strict 2PL Strict Two-phase Locking (Strict 2PL) Protocol: Each Xact must obtain a S (shared) lock on object before reading, and an X (exclusive) lock on object before writing. All locks held by a transaction are released when the transaction completes If an Xact holds an X lock on an object, no other Xact can get a lock (S or X) on that object. Strict 2PL allows only schedules whose precedence graph is acyclic (conflict serializable). Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 12 6

7 Two-Phase Locking (2PL) Two-Phase Locking Protocol Each Xact must obtain a S (shared) lock on object before reading, and an X (exclusive) lock on object before writing. Release locks anytime (before the end commit or abort). A transaction can not request additional locks once it releases any locks. If an Xact holds an X lock on an object, no other Xact can get a lock (S or X) on that object. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 13 2PL Every transaction has a growing phase in which it acquires locks, followed by a shrinking phase in which it releases locks. 2PL allows only schedules whose precedence graph is acyclic (conflict serializable). Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 14 7

8 Strict Schedule A schedule is strict if a value written by a transaction T is not read or overwritten by other transactions until T either aborts or commits. Strict schedules are recoverable, cascading aborts are not required. Strict 2PL improves on 2PL by guaranteeing that every allowed schedule is strict in addition to being conflict serializable. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 15 Lock Management Lock and unlock requests are handled by the lock manager Lock table entry: Number of transactions currently holding a lock Type of lock held (Shared or exclusive) Pointer to queue of lock requests Locking and unlocking have to be atomic operations Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 16 8

9 Handle Lock Request If a Shared lock is requested, the queue of requests is empty, and the object is not locked in exclusive mode, the lock is granted and the lock table entry is updated. If an exclusive lock is requested, the queue is empty, and the object is not locked in either modes, the lock is granted and the entry is updated. Otherwise, the lock request is queued. The transaction is suspended. The request queue will prevent starving a request. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 17 Unlock When a transaction aborts or commits, it releases all its locks. Lock manager updates table entry and examine the request queue. All shared locks requested could be granted at the same time. The suspended transaction is woken up and continues. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 18 9

10 Case Study If T1 has a shared lock on object O, T2 requests an exclusive lock, T2 s request is queued. If T3 requests a shared lock after T2, its request could not be granted right away even its request is compatible with shared lock by T1. T3 s request is queued. T2 will get the lock before T3 (not starving). Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 19 Lock Conversions Method 1: The request to upgrade a Shared lock to an exclusive lock must be handled specially. May cause deadlock Method 2: An exclusive lock is obtained initially and downgraded to a Shared lock if it is sufficient. Reduce concurrency, improves throughput. Method 3: A third type lock, an update lock, is granted initially. It could be upgraded or downgraded later. Improve concurrency. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 20 10

11 Deadlocks Deadlock: Cycle of transactions waiting for locks to be released by each other. Two ways of dealing with deadlocks: Deadlock prevention Deadlock detection Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 21 Deadlock Detection Create a waits-for graph: Nodes are transactions There is an edge from Ti to Tj if Ti is waiting for Tj to release a lock Tj read-locks X then Ti tries to write-lock it Tj write-locks X then Ti tries to read-lock it Tj write-locks X then Ti tries to write-lock it Periodically check for cycles in the waits-for graph Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 22 11

12 Deadlock Detection (Continued) Example: T1: S(A), R(A), S(B) T2: X(B),W(B) X(C) T3: S(C), R(C) X(A) T4: X(B) T1 T2 T1 T2 T4 T3 T4 T3 Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 23 Example T1 T2 T3 R(X) R(Y) W(X) R(X) T1 T3 T2 Wait-For Graph R(Y) R(Z) W(Z) R(X) T1 T3 T2 W(Y) Precedence Graph Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 24 12

13 Solve a Deadlock When a deadlock is detected, a transaction should be aborted and all its locks are released. Criteria to follow: The one with the fewest locks. The one has done the least work. The one is farthest from completion. The one has been repeated restarted, it should be favored during deadlock detection. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 25 A Simple Alternative Method Instead of maintaining a waits-for graph, a timeout mechanism is adopted. A transaction has been waiting for too long for a lock, we assume that it is in a deadlock cycle. So the transaction is aborted. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 26 13

14 Deadlock Prevention Assign priorities based on timestamps. The oldest transaction has the highest priority. Assume Ti wants a lock that Tj holds. The lock manager uses one of two possible policies: Wait-Die: Lower one dies. If Ti has higher priority, Ti waits for Tj; otherwise Ti aborts Wound-wait: Lower one waits. If Ti has higher priority, Tj aborts; otherwise Ti waits Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 27 Two Schemes Wait-Die is non-preemptive only a transaction requesting a lock can be aborted. A low-priority transaction never waits for a higher one. Disadvantage: A younger transaction that conflicts with an older one may be repeatedly aborted. Advantage: A transaction that has all locks does not abort. Wound-wait is preemptive. A high-priority transaction never waits for a lower one. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 28 14

15 Deadlock Prevention In both schemes, a higher-priority transaction is never aborted. If a transaction re-starts, make sure it has its original timestamp. Thus, it has higher priority. It is ensured that no transaction is perennially aborted because it never has a sufficiently high priority. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 29 Dynamic Databases If we relax the assumption that the DB is a fixed collection of objects, even Strict 2PL will not assure serializability: T1 locks all pages containing sailor records with rating = 1, and finds oldest sailor (say, age = 71). Next, T2 inserts a new sailor; rating = 1, age = 96. T2 also deletes oldest sailor with rating = 2 (and, say, age = 80), and commits. T1 now locks all pages containing sailor records with rating = 2, and finds oldest (say, age = 63). No consistent DB state where T1 is correct! Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 30 15

16 The Phantom Problem T1 implicitly assumes that it has locked the set of all sailor records with rating = 1. Assumption only holds if no sailor records are added while T1 is executing! Need some mechanism to enforce this assumption. (Index locking and predicate locking.) Example shows that conflict serializability guarantees serializability only if the set of objects is fixed! Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 31 Index Locking Index r=1 Data If there is a dense index on the rating field using Alternative (2), T1 should lock the index page containing the data entries with rating = 1. If there are no records with rating = 1, T1 must lock the index page where such a data entry would be, if it existed! If there is no suitable index, T1 must lock all pages, and lock the file/table to prevent new pages from being added, to ensure that no new records with rating = 1 are added. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 32 16

17 Predicate Locking Grant lock on all records that satisfy some logical predicate, e.g. age > 2*salary. Index locking is a special case of predicate locking for which an index supports efficient implementation of the predicate lock. What is the predicate in the sailor example? In general, predicate locking has a lot of locking overhead. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 33 Locking in B+ Trees How can we efficiently lock a particular leaf node? One solution: Ignore the tree structure, just lock pages while traversing the tree, following 2PL. This has terrible performance! Root node (and many higher level nodes) become bottlenecks because every tree access begins at the root. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 34 17

18 Two Useful Observations Higher levels of the tree only direct searches for leaf pages. For inserts, a node on a path from root to modified leaf must be locked (in X mode, of course) only if a split can propagate up to it from the modified leaf. (Similar point holds w.r.t. deletes.) We can exploit these observations to design efficient locking protocols that guarantee serializability. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 35 A Simple Tree Locking Algorithm Search: Start at root and go down; repeatedly, S lock child then unlock parent. Insert/Delete: Start at root and go down, obtaining X locks as needed. Once child is locked, check if it is safe: If child is safe, release all locks on ancestors. Safe node: Node such that changes will not propagate up beyond this node. Inserts: Node is not full. Deletes: Node is not half-empty. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 36 18

19 ROOT Example 20 A 35 B Do: 1) Search 38* 2) Delete 38* 3) Insert 45* 4) Insert 25* 23 F C 20* G H I D 22* 23* 24* 35* 36* 38* 41* 44* E Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 37 Example One Search data entry 38*, Ti holds S lock on node A, reads it and decides to exam node B. S lock on node B and releases lock on node A, reads it and decides to exam node C. S lock on node C and releases lock on node B, reads it and decides to exam node D. S lock on node D and releases lock on node C. A transaction always holds a lock on one node in the path, to force other transactions to wait. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 38 19

20 Example Two Insert data entry 45*, Ti holds S lock on node A, reads it and decides to exam node B. S lock on node B, reads it, and releases lock on node A, and decides to exam node C. S lock on node C, reads it, and dose not release lock on node B since C is full, and decides to exam node E. X lock on node E, reads it, and release locks on C and then B since node E has space for new entry. Insert 45* on node E. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 39 Example Three Insert data entry 25*, Ti holds S lock on node A, reads it and decides to exam node B. S lock on node B, reads it, and releases lock on node A, and decides to exam node F. S lock on node F, reads it, and releases lock on node B since it is not full, and decides to exam node H. X lock on node H; reads it, since it is full and must be split, and its parent node F should be modified, too. S lock on F is requested to be updated to X lock. If no other transaction holds S lock on F, the upgrade is granted. Node H is split to insert 25*. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 40 20

21 Optimistic CC (Kung-Robinson) Locking is a conservative approach in which conflicts are prevented. Disadvantages: Lock management overhead. Deadlock detection/resolution. Lock contention for heavily used objects. If conflicts are rare, we might be able to gain concurrency by not locking, and instead checking for conflicts before Xacts commit. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 41 Kung-Robinson Model Xacts have three phases: READ: Xacts read from the database, but make changes to private copies of objects. VALIDATE: Check for conflicts. Abort if there is conflict, and clear private copies. WRITE: Make local copies of changes public. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 42 21

22 Validation Test conditions that are sufficient to ensure that no conflict occurs. Each Xact is assigned a timestamp (TS) at end of READ phase, just before validation begins. ReadSet(Ti): Set of objects read by Xact Ti. WriteSet(Ti): Set of objects modified by Ti. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 43 Test 1 For all i and j such that TS(Ti) < TS(Tj), check that Ti completes before Tj begins. Ti R V W Tj R V W Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 44 22

23 Test 2 For all i and j such that TS(Ti) < TS(Tj), check that: Ti completes before Tj begins its Write phase + The intersection of WriteSet(Ti) and ReadSet(Tj) is empty. Ti R V W R V W Tj Does Tj read dirty data? Does Ti overwrite Tj s writes? Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 45 Test 3 For all i and j such that TS(Ti) < TS(Tj), check that: Ti completes Read phase before Tj completes Read + The intersection of WriteSet(Ti) and ReadSet(Tj) is empty + The intersection of WriteSet(Ti) and WriteSet(Tj) is empty. Ti R V W R V W Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 46 Tj Does Tj read dirty data? Does Ti overwrite Tj s writes? 23

24 Overheads in Optimistic CC Must record read/write activity in ReadSet and WriteSet per Xact. Must create and destroy these sets as needed. Must check for conflicts during validation, and must make validated writes ``global. Critical section can reduce concurrency. Optimistic CC restarts Xacts that fail validation. Work done so far is wasted; requires clean-up. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 47 Summary There are several lock-based concurrency control schemes (Strict 2PL, 2PL). Conflicts between transactions can be detected in the dependency graph The lock manager keeps track of the locks issued. Deadlocks can either be prevented or detected. Naïve locking strategies may have the phantom problem Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 48 24

25 Summary (Contd.) Index locking is common, and affects performance significantly. Needed when accessing records via index. Needed for locking logical sets of records (index locking/predicate locking). Tree-structured indexes: Straightforward use of 2PL very inefficient. Bayer-Schkolnick illustrates potential for improvement. In practice, better techniques now known; do record-level, rather than page-level locking. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 49 Summary (Contd.) Optimistic CC aims to minimize CC overheads in an ``optimistic environment where reads are common and writes are rare. Optimistic CC has its own overheads however; most real systems use locking. Database Management Systems 3ed, R. Ramakrishnan and J. Gehrke 50 25