TRANSACTIONS (AND EVENT-DRIVEN PROGRAMMING) EE324

Transcription

1 TRANSACTIONS (AND EVENT-DRIVEN PROGRAMMING) EE324

2 Concurrency Control 2 General organization of managers for handling transactions.

3 Two-phase locking is pessimistic 3 Acts to prevent non-serializable schedules from arising: pessimistically ass umes conflicts are fairly likely Can deadlock, e.g. T1 reads x then writes y; T2 reads y then writes x. T his doesn t always deadlock but it is capable of deadlocking Overcome by aborting if we wait for too long, Or by designing transactions to obtain locks in a known and agreed upon ordering

4 Transactions T and U with exclusive locks 4 Transaction T: balance = b.getbalance() b.setbalance(bal*1.1) a.withdraw(bal/10) Transaction U: balance = b.getbalance() b.setbalance(bal*1.1) c.withdraw(bal/10) Operations Locks Operations Locks opentransaction bal = b.getbalance() lock B b.setbalance(bal*1.1) opentransaction a.withdraw(bal/10) lock A bal = b.getbalance() waits for T s lock on B closetransaction unlock A, B lock B b.setbalance(bal*1.1) c.withdraw(bal/10) closetransaction lock C unlock B, C

5 Schemes for Concurrency control 5 Locking Optimistic concurrency control (CDK516.5) Time-stamp based concurrency control (not going to cover)

6 Optimistic Concurrency Control 6 Drawbacks of locking Overhead of lock maintenance Deadlocks Reduced concurrency Optimistic Concurrency Control In most applications, likelihood of conflicting accesses by concurrent transacti ons is low Transactions proceed as though there are no conflicts

7 Optimistic Concurrency Control 7 Three phases: Working Phase transactions read and write private copies of objects ( most recently committed) Validation Phase Once transaction is done, the transaction is validated to establish whether or not its operations on objects conflict with operation s of other transactions on the same object. If not conflict, can commit; else some form of conflict resolution is needed and the transaction may abort. Update Phase if commit, private copies are used to make permanent c hange.

8 Validation of transactions 8 Working Validation Update T 1 Earlier committed transactions T 2 T 3 Transaction being validated T v active 1 Later active transactions active 2

9 Validation conflict detection Tv Ti Rule Write Read Ti must not read objects written by Tv Read Write Tv must not read objects written by Ti Write Write Ti must not write objects written by Tv and Tv must not write objects written by Ti

10 Optimistic concurrency control Transactions are numbered. Validation and update occurs in side the critical section. (Satisfies rule 3) Backward validation Rule 1 is satisfied because all read operations of earlier overlapping transactions were performed before the validation of Tv started; they cannot be affected by Tv s write. Read set of Tv must be compared with the write sets of T2 and T3. (Rule 2) In other words, the read set of the transaction being validated is compared with the write set of other overlapping transactions that have already committed. Thus, we need to keep the write set of T2 and T3 for a while after their commit.

11 Today's Lecture 11 Distributed transactions

12 Distributed Transactions 12 Motivation Provide distributed atomic operations at multiple servers that maintain share d data for clients Provide recoverability from server crashes Properties Atomicity, Consistency, Isolation, Durability (ACID) Concepts: commit, abort, distributed commit

13 Transactions in distributed systems 13 Main issue that arises is that now we can have multiple database serve rs that are touched by one transaction Reasons? Data spread around: each owns subset Could have replicated some data object on multiple servers, e.g. to load-ba lance read access for large client set Might do this for high availability

14 Atomic Commit Protocols 14 The atomicity of a transaction requires that when a distributed transact ion comes to an end, either all of its operations are carried out or non e of them One phase commit Coordinator tells all participants (servers) to commit Keep on repeating it until all participants reply If a participant cannot commit (say because of concurrency control), no way to infor m coordinator. Also, no way for the coordinator to abort.

15 The two-phase commit protocol (2PC) Designed to allow any participant to abort its part of transaction But this means, the whole transaction must be aborted Why? First phase: all participants vote (abort or commit). If voted commit, make all changed permanent (durability) and go to prepared state. Log this fact. Participants will eventually commit (if the coordinator says so) even it crashes. Second phase: Joint decision

16 The two-phase commit protocol Phase 1 (voting phase): 1. The coordinator sends a cancommit? (VOTE_REQUEST) request to each of the participants in the transaction. 2. When a participant receives a cancommit? request it repli es with its vote Yes (VOTE_COMMIT) or No (VOTE_AB ORT) to the coordinator. Before voting Yes, it prepares to commit by saving objects in permanent storage. If the vot e is No the participant aborts immediately.

17 The two-phase commit protocol Phase 2 (completion according to outcome of vote): 3. The coordinator collects the votes (including its own). (a)if there are no failures and all the votes are Yes the coordinator d ecides to commit the transaction and sends a docommit (GLOBA L_COMMIT) request to each of the participants. (b)otherwise the coordinator decides to abort the transaction and se nds doabort (GLOBAL_ABORT) requests to all participants that voted Yes. 4. Participants that voted Yes are waiting for a docommit or doabort reques t from the coordinator. When a participant receives one of these messag es it acts accordingly and in the case of commit, makes a havecommitte d call as confirmation to the coordinator.

18 Communication in two-phase commit protocol 18 Coordinator Participant step 1 status prepared to commit (waiting for votes) cancommit? Yes step 2 status prepared to commit 3 committed docommit (uncertain) havecommitted 4 committed done

19 Commit protocol illustrated 19 ok to commit?

20 Commit protocol illustrated 20 ok to commit? commit ok with us

21 Operations for two-phase commit protocol 21 cancommit?(trans)-> Yes / No Call from coordinator to participant to ask whether it can commit a transaction. Participa nt replies with its vote. docommit(trans) Call from coordinator to participant to tell participant to commit its part of a transaction. doabort(trans) Call from coordinator to participant to tell participant to abort its part of a transaction. havecommitted(trans, participant) Call from participant to coordinator to confirm that it has committed the transaction. getdecision(trans) -> Yes / No Call from participant to coordinator to ask for the decision on a transaction after it has vo ted Yes but has still had no reply after some delay. Used to recover from server crash or d elayed messages.

22 Two-Phase Commit protocol 3 (TV sec 8.5) 22 actions by coordinator: while START _2PC to local log; multicast VOTE_REQUEST to all participants; while not all votes have been collected { wait for any incoming vote; if timeout { write GLOBAL_ABORT to local log; multicast GLOBAL_ABORT to all participants; exit; } record vote; } if all participants sent VOTE_COMMIT and coordinator votes COMMIT{ write GLOBAL_COMMIT to local log; multicast GLOBAL_COMMIT to all participants; } else { write GLOBAL_ABORT to local log; multicast GLOBAL_ABORT to all participants; }

23 Two-Phase Commit protocol actions by participant: write INIT to local log; wait for VOTE_REQUEST from coordinator; if timeout { write VOTE_ABORT to local log; exit; } if participant votes COMMIT { write VOTE_COMMIT to local log; send VOTE_COMMIT to coordinator; wait for DECISION from coordinator; if timeout { multicast DECISION_REQUEST to other participants; wait until DECISION is received; /* remain blocked */ write DECISION to local log; } if DECISION == GLOBAL_COMMIT write GLOBAL_COMMIT to local log; else if DECISION == GLOBAL_ABORT write GLOBAL_ABORT to local log; } else { write VOTE_ABORT to local log; send VOTE ABORT to coordinator; }

24 Two-Phase Commit protocol a) The finite state machine for the coordinator in 2PC. b) The finite state machine for a participant. Coordinator and participants have blocking state. When a fa ilure occurs, other process may be indefinitely waiting. There needs to be a timeout mechanism.

25 Two Phase Commit Protocol Timeouts Wait in Coordinator use a time-out mechanism to detect participant crashes. Send GLOBAL_ABORT Init in Participant Can also use a time-out and send VOTE_ABORT Ready in Participant P abort is not an option (since already voted to COM MIT and so coordinator might eventually send GLOBAL_COMMIT). Can contac t another participant Q and choose an action based on its state. State of Q COMMIT ABORT INIT READY Action by P Transition to COMMIT Transition to ABORT Both P and Q transition to ABORT (Q sends VOTE_ABORT) Contact more participants. If all participants are READY, must wait for coordinator to recover

26 Two Phase Commit Protocol - 7 Recovery To ensure that a process can actually recover, it must save its state to persistent storage. If a participant was in INIT (before crash), it can safely decide to locally abort when it recovers and inform the coordinator. If it was COMMIT and ABORT, retransmit its decision to the coordinator. If it was READY, contact other participant Q (Send DECISION_REQUEST), similar to the timeout situation.

27 Two-Phase Commit protocol actions for handling decision requests: /* executed by separate thread */ while true { wait until any incoming DECISION_REQUEST is received; /* rem ain blocked */ read most recently recorded STATE from the local log; if STATE == GLOBAL_COMMIT send GLOBAL_COMMIT to requesting participant; else if STATE == INIT or STATE == GLOBAL_ABORT send GLOBAL_ABORT to requesting participant; else skip; /* participant remains blocked */

28 Three Phase Commit protocol Problem with 2PC If coordinator crashes, participants cannot reach a decision, stay blocked unt il coordinator recovers Three Phase Commit (3PC): proof in [SS 1983] There is no single state from which it is possible to make a transition directly to either COMMIT or ABORT states There is no state in which it is not possible to make a final decision, and from which a transition to COMMIT can be made

29 Three-Phase Commit protocol a) Finite state machine for the coordinator in 3PC b) Finite state machine for a participant

30 Three Phase Commit Protocol Recovery Wait in Coordinator same Init in Participant same PreCommit in Coordinator Some participant has crashed but we know it wanted to commit. GLOBAL_COMMIT the application knowing that once the participant recovers, it will commit. Ready or PreCommit in Participant P (i.e. P has voted to COMMIT) State of Q PRECOMMIT ABORT INIT READY Action by P Transition to PRECOMMIT. If all participants in PRECOMMIT and form a majority, then COMMIT the transaction Transition to ABORT Both P (in READY) and Q transition to ABORT (Q sends VOTE_ABORT). It can be shown that no other participants can be in PRECOMMIT Contact more participants. If can contact a majority and they are in Ready, then ABORT the transaction. If the participants contacted in PreCommit it is safe to COMMIT the transaction Note: if any participant is in state PRECOMMIT, it is impossible for any other participant to be in any state other than READY or PRECOMMIT.

31 Things we have learned so far ACID Concurrency control Distributed atomic commit

32 Two Views of Distributed Systems 32 Optimist: A distributed system is a collection of independent computer s that appears to its users as a single coherent system Pessimist: You know you have one when the crash of a computer you ve never heard of stops you from getting any work done. (Lamport)

33 Recurring Theme 33 Academics like: Clean abstractions Strong semantics Things that prove they are smart Users like: Systems that work (most of the time) Systems that scale Consistency per se isn t important Eric Brewer had the following observations

34 A Clash of Cultures 34 Classic distributed systems: focused on ACID semantics (transaction seman tics) Atomicity: either the operation (e.g., write) is performed on all replicas or is no t performed on any of them Consistency: after each operation all replicas reach the same state Isolation: no operation (e.g., read) can see the data from another operation (e. g., write) in an intermediate state Durability: once a write has been successful, that write will persist indefinitely Modern Internet systems: focused on BASE Basically Available Soft-state (or scalable) Eventually consistent

35 ACID vs BASE ACID BASE Strong consistency for transacti ons highest priority Availability less important Pessimistic Rigorous analysis Complex mechanisms Availability and scaling highest priorities Weak consistency Optimistic Best effort Simple and fast 35

36 Why Not ACID+BASE? 36 What goals might you want from a system? C, A, P Strong Consistency: all clients see the same view, even in the presence of u pdates High Availability: all clients can find some replica of the data, even in the presence of failures Partition-tolerance: the system properties hold even when the system is part itioned

37 CAP Theorem [Brewer] 37 You can only have two out of these three properties The choice of which feature to discard determines the nature of your s ystem

38 Consistency and Availability 38 Comment: Providing transactional semantics requires all functioning nodes to be in contact with each other (no partition) Examples: Single-site and clustered databases Other cluster-based designs Typical Features: Two-phase commit Cache invalidation protocols Classic DS style

39 Partition-Tolerance and Availability 39 Comment: Once consistency is sacrificed, life is easy. Examples: DNS Web caches Practical distributed systems for mobile environments: Coda, Bayou Typical Features: Optimistic updating with conflict resolution This is the Internet design style TTLs and lease cache management

40 Voting with their Clicks 40 In terms of large-scale systems, the world has voted with their clicks: Consistency less important than availability and partition-tolerance