' & $ Chapter 18: Distributed Databases · Distributed Data Storage · Network Transparency · Distributed Query Processing · Distributed Transaction Model · Commit Protocols · Coordinator Selection · Concurrency Control · Deadlock Handling · Multidatabase Systems Database Systems Concepts 18.1 Silberschatz, Korth and Sudarshan c 1997 ' & $ Distributed Database System · Database is stored on several computers that communicate via media such as wide-area networks, telephone lines, or local area networks. · Appears to user as a single system · Processes complex queries · Processing may be done at a site other than the initiator of the request · Transaction management · Optimization of queries provided automatically Database Systems Concepts 18.2 Silberschatz, Korth and Sudarshan c 1997 ' & $ Distributed Data Storage Assume relational data model · Replication: system maintains multiple copies of data, stored in different sites, for faster retrieval and fault tolerance. · Fragmentation: relation is partitioned into several fragments stored in distinct sites. · Replication and fragmentation: relation is partitioned into several fragments; system maintains several identical replicas of each such fragment. Database Systems Concepts 18.3 Silberschatz, Korth and Sudarshan c 1997 ' & $ Data Replication · A relation or fragment of a relation is replicated if it is stored redundantly in two or more sites. · Full replication of a relation is the case where the relation is stored at all sites. · Fully redundant databases are those in which every site contains a copy of the entire database. Database Systems Concepts 18.4 Silberschatz, Korth and Sudarshan c 1997 ' & $ Data Replication (Cont.) · Advantages of Replication ­ Availability: failure of a site containing relation r does not result in unavailability of r if replicas exist. ­ Parallelism: queries on r may be processed by several nodes in parallel. ­ Reduced data transfer: relation r is available locally at each site containing a replica of r. · Disadvantages of Replication ­ Increased cost of updates: each replica of relation r must be updated. ­ Increased complexity of concurrency control: concurrent updates to distinct replicas may lead to inconsistent data unless special concurrency control mechanisms are implemented. Database Systems Concepts 18.5 Silberschatz, Korth and Sudarshan c 1997 ' & $ Data Fragmentation · Division of relation r into fragments r1, r2, ..., rn which contain sufficient information to reconstruct relation r. · Horizontal fragmentation: each tuple of r is assigned to one or more fragments. · Vertical fragmentation: the schema for relation r is split into several smaller schemas. ­ All schemas must contain a common candidate key (or superkey) to ensure lossless join property. ­ A special attribute, the tuple-id attribute may be added to each schema to serve as a candidate key. · Fragments may be successively fragmented to an arbitrary depth. Vertical and horizontal fragmentation can be mixed. · Example: relation account with following schema Account-schema = (branch-name, account-number, balance) Database Systems Concepts 18.6 Silberschatz, Korth and Sudarshan c 1997 ' & $ Horizontal Fragmentation of account Relation branch-name account-number balance Hillside A-305 500 Hillside A-226 336 Hillside A-155 62 account1 branch-name account-number balance Valleyview A-177 205 Valleyview A-402 10000 Valleyview A-408 1123 Valleyview A-639 750 account2 Database Systems Concepts 18.7 Silberschatz, Korth and Sudarshan c 1997 ' & $ Vertical Fragmentation of deposit Relation branch-name customer-name tuple-id Hillside Lowman 1 Hillside Camp 2 Valleyview Camp 3 Valleyview Kahn 4 Hillside Kahn 5 Valleyview Kahn 6 Valleyview Green 7 deposit1 account-number balance tuple-id A-305 500 1 A-226 336 2 A-177 205 3 A-402 10000 4 A-155 62 5 A-408 1123 6 A-639 750 7 deposit2 Database Systems Concepts 18.8 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advantages of Fragmentation · Horizontal: ­ allows parallel processing on a relation ­ allows a global table to be split so that tuples are located where they are most frequently accessed · Vertical: ­ allows for further decomposition than can be achieved with normalization ­ tuple-id attribute allows efficient joining of vertical fragments ­ allows parallel processing on a relation ­ allows tuples to be split so that each part of the tuple is stored where it is most frequently accessed Database Systems Concepts 18.9 Silberschatz, Korth and Sudarshan c 1997 ' & $ Network Transparency · Degree to which system users may remain unaware of the details of how and where the data items are stored in a distributed system · Consider transparency issues in relation to: ­ Naming of data items ­ Replication of data items ­ Fragmentation of data items ­ Location of fragments and replicas Database Systems Concepts 18.10 Silberschatz, Korth and Sudarshan c 1997 ' & $ Naming of Data Items ­ Criteria 1. Every data item must have a system-wide unique name. 2. It should be possible to find the location of data items efficiently. 3. It should be possible to change the location of data items transparently. 4. Each site should be able to create new data items autonomously. Database Systems Concepts 18.11 Silberschatz, Korth and Sudarshan c 1997 ' & $ Centralized Scheme -- Name Server · Structure: ­ name server assigns all names ­ each site maintains a record of local data items ­ sites ask name server to locate non-local data items · Advantages: ­ satisfies naming criteria 1-3 · Disadvantages: ­ does not satisfy naming criterion 4 ­ name server is a potential performance bottleneck ­ name server is a single point of failure Database Systems Concepts 18.12 Silberschatz, Korth and Sudarshan c 1997 ' & $ Use of Aliases · Alternative to centralized scheme: each site prefixes its own site identifier to any name that it generates, i.e., site17.account. ­ Fulfills having a unique identifier, and avoids problems associated with central control. ­ However, fails to achieve network transparency. · Solution: Create a set of aliases for data items; Store the mapping of aliases to the real names at each site. · The user can be unaware of the physical location of a data item, and is unaffected if the data item is moved from one site to another. Database Systems Concepts 18.13 Silberschatz, Korth and Sudarshan c 1997 ' & $ Use of Aliases (Cont.) · Each replica and each fragment of a data item must have a unique name. ­ Use postscripts to determine those replicas that are replicas of the same data item, and those fragments that are fragments of the same data item. ­ fragments of same data item: ".f1", ".f2", . . . , ".fn" ­ replicas of same data item: ".r1", ".r2", . . . , ".rn" site17.account.f3.r2 refers to replica 2 of fragment 3 of account; this item was generated by site 17. Database Systems Concepts 18.14 Silberschatz, Korth and Sudarshan c 1997 ' & $ Name-Translation Algorithm if name appears in the alias table then expression := map (name) else expression := name; function map (n) if n appears in the replica table then result := name of replica of n; if n appears in the fragment table then begin result := expression to construct fragment; for each n in result do begin replace n in result with map (n ); end end return result; Database Systems Concepts 18.15 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example of Name-Translation Scheme · A user at the Hillside branch, (site S1), uses the alias local-account for the local fragment account.f1 of the account relation. · When this user references local-account, the query-processing subsystem looks up local-account in the alias table, and replaces local-account with S1.account.f1. · If S1.account.f1 is replicated, the system must consult the replica table in order to choose a replica. · If this replica is fragmented, the system must examine the fragmentation table to find out how to reconstruct the relation. · Usually only need to consult one or two tables, however, the algorithm can deal with any combination of successive replication and fragmentation of relations. Database Systems Concepts 18.16 Silberschatz, Korth and Sudarshan c 1997 ' & $ Transparency and Updates · Must ensure that all replicas of a data item are updated and that all affected fragments are updated. · Consider the account relation and the insertion of the tuple: ("Valleyview", A-733, 600) · Horizontal fragmentation of account account1 = branch-name = "Hillside" (account) account2 = branch-name = "Valleyview" (account) ­ Predicate Pi is associated with the ith fragment ­ Apply Pi to the tuple ("Valleyview", A-733, 600) to test whether that tuple must be inserted in the ith fragment ­ Tuple inserted into account2 Database Systems Concepts 18.17 Silberschatz, Korth and Sudarshan c 1997 ' & $ Transparency and Updates (Cont.) · Vertical fragmentation of deposit into deposit1 and deposit2 · The tuple ("Valleyview", A-733, `Jones", 600) must be split into two fragments: ­ one to be inserted into deposit1 ­ one to be inserted into deposit2 · If deposit is replicated, the tuple ("Valleyview", A-733, "Jones"600) must be inserted in all replicas · Problem: If deposit is accessed concurrently it is possible that one replica will be updated earlier than another (see section on Concurrency Control). Database Systems Concepts 18.18 Silberschatz, Korth and Sudarshan c 1997 ' & $ Distributed Query Processing · For centralized systems, the primary criterion for measuring the cost of a particular strategy is the number of disk accesses. · In a distributed system, other issues must be taken into account: ­ The cost of data transmission over the network. ­ The potential gain in performance from having several sites process parts of the query in parallel. Database Systems Concepts 18.19 Silberschatz, Korth and Sudarshan c 1997 ' & $ Query Transformation · Translating algebraic queries to queries on fragments. ­ It must be possible to construct relation r from its fragments ­ Replace relation r by the expression to construct relation r from its fragments · Site selection for query processing. Database Systems Concepts 18.20 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example Query · Consider the horizontal fragmentation of the account relation into account1 = branch-name = "Hillside" (account) account2 = branch-name = "Valleyview" (account) · The query branch-name = "Hillside" (account) becomes branch-name = "Hillside" (account1 account2) which is optimized into branch-name = "Hillside" (account1) branch-name = "Hillside" (account2) Database Systems Concepts 18.21 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example Query (Cont.) · Since account1 has only tuples pertaining to the Hillside branch, we can eliminate the selection operation. · Apply the definition of account2 to obtain branch-name = "Hillside" (branch-name = "Valleyview" (account)) · This expression is the empty set regardless of the contents of the account relation. · Final strategy is for the Hillside site to return account1 as the result of the query. Database Systems Concepts 18.22 Silberschatz, Korth and Sudarshan c 1997 ' & $ Simple Join Processing Consider the following relational algebra expression in which the three relations are neither replicated nor fragmented account 1 depositor 1 branch · account is stored at site S1 · depositor at S2 · branch at S3 · For a query issued at site SI, the system needs to produce the result at site SI. Database Systems Concepts 18.23 Silberschatz, Korth and Sudarshan c 1997 ' & $ Possible Query Processing Strategies · Ship copies of all three relations to site SI and choose a strategy for processing the entire query locally at site SI. · Ship a copy of the account relation to site S2 and compute temp1 = account 1 depositor at S2. Ship temp1 from S2 to S3, and compute temp2 = temp1 1 branch at S3. Ship the result temp2 to SI. · Devise similar strategies, exchanging the roles of S1, S2, S3. · Must consider following factors: ­ amount of data being shipped ­ cost of transmitting a data block between sites ­ relative processing speed at each site Database Systems Concepts 18.24 Silberschatz, Korth and Sudarshan c 1997 ' & $ Semijoin Strategy · Let r1 be a relation with schema R1 stored at site S1 Let r2 be a relation with schema R2 stored at site S2 · Evaluate the expression r1 1 r2, and obtain the result at S1. 1. Compute temp1 R1 R2 (r1) at S1. 2. Ship temp1 from S1 to S2. 3. Compute temp2 r2 1 temp1 at S2. 4. Ship temp2 from S2 to S1. 5. Compute r1 1 temp2 at S1. This is the result of r1 1 r2. Database Systems Concepts 18.25 Silberschatz, Korth and Sudarshan c 1997 ' & $ Formal Definition · The semijoin of r1 with r2, is denoted by: r1< r2 it is defined by: R1 (r1 1 r2) · Thus, r1 < r2 selects those tuples of r1 that contributed to r1 1 r2. · In step 3 above, temp2 = r2 < r1. · For joins of several relations, the above strategy can be extended to a series of semijoin steps. Database Systems Concepts 18.26 Silberschatz, Korth and Sudarshan c 1997 ' & $ Join Strategies that Exploit Parallelism · Consider r1 1 r2 1 r3 1 r4 where relation ri is stored at site Si . The result must be presented at site S1. · Pipelined-join strategy ­ r1 is to S2 and r1 1 r2 is computed at S2; simultaneously r3 is shipped to S4 and r3 1 r4 is computed at S4 ­ S2 ships tuples of (r1 1 r2) to S1 as they are produced; S4 ships tuples of (r3 1 r4) to S1 ­ Once tuples of (r1 1 r2) and (r3 1 r4) arrive at S1, (r1 1 r2) 1 (r3 1 r4) is computed in parallel with the computation of (r1 1 r2) at S2 and the computation of (r3 1 r4) at S4. Database Systems Concepts 18.27 Silberschatz, Korth and Sudarshan c 1997 ' & $ Distributed Transaction Model · Transactions may access data at several sites · Each site has a local transaction manager responsible for: ­ Maintaining a log for recovery purposes. ­ Participating in coordinating the concurrent execution of the transactions executing at that site. · Each site has a transaction coordinator, which is responsible for: ­ Starting the execution of transactions that originate at the site. ­ Distributing subtransactions to appropriate sites for execution. ­ Coordinating the termination of each transaction that originates at the site, which may result in the transaction being committed at all sites or aborted at all sites. Database Systems Concepts 18.28 Silberschatz, Korth and Sudarshan c 1997 ' & $ Transaction System Architecture TM1 TMn computer 1 computer n TC1 TCn transaction coordinator transaction manager Database Systems Concepts 18.29 Silberschatz, Korth and Sudarshan c 1997 ' & $ System Failure Modes · Failures unique to distributed systems: ­ Failure of a site. ­ Loss of messages. ­ Failure of a communication link. ­ Network partition. · The configurations of how sites are connected physically can be compared in terms of: ­ Installation cost. ­ Communication cost. ­ Availability. Database Systems Concepts 18.30 Silberschatz, Korth and Sudarshan c 1997 ' & $ System Failure Modes (Cont.) · Partially connected networks have direct links between some, but not all, pairs of sites. ­ Lower installation cost than fully connected network ­ Higher communication cost to route messages between two sites that are not directly connected Database Systems Concepts 18.31 Silberschatz, Korth and Sudarshan c 1997 ' & $ Network Topology A EB C FD A B E D CF A B E D CF A B E D C F C E F D B A fully connected network tree structured network ring network partially connected network star network Database Systems Concepts 18.32 Silberschatz, Korth and Sudarshan c 1997 ' & $ Network Topology (Cont.) · A partitioned system is split into two (or more) subsystems (partitions) that lack any connection. · Tree-structured: low installation and communication costs; the failure of a single link can partition network · Ring: At least two links must fail for partition to occur; communication cost is high · Star: ­ the failure of a single link results in a network partition, but since one of the partitions has only a single site it can be treated as a single-site failure ­ low communication cost ­ failure of the central site results in every site in the system becoming disconnected Database Systems Concepts 18.33 Silberschatz, Korth and Sudarshan c 1997 ' & $ Robustness · A robust system must: ­ Detect site or link failures ­ Reconfigure the system so that computation may continue. ­ Recover when a processor or link is repaired. · Handling failure types: ­ Retransmit lost messages. ­ Unacknowledged retransmits indicate link failure; find alternative route for message. ­ Failure to find alternative route is a symptom of network partition. · Network link failures and site failures are generally indistinguishable. Database Systems Concepts 18.34 Silberschatz, Korth and Sudarshan c 1997 ' & $ Procedure to Reconfigure System · If replicated data is stored at the failed site, update the catalog so that queries do not reference the copy at the failed site. · Transactions active at the failed site should be aborted. · If the failed site is a central server for some subsystem, an election must be held to determine the new server. · Reconfiguration scheme must work correctly in case of network partitioning; must avoid: ­ Electing two or more central servers in distinct partitions. ­ Updating replicated data item by more than one partition · Represent recovery tasks as a series of transactions; concurrent control subsystem and transaction management subsystem may then be relied upon for proper reintegration. Database Systems Concepts 18.35 Silberschatz, Korth and Sudarshan c 1997 ' & $ Commit Protocols · Commit protocols are used to ensure atomicity across sites ­ a transaction which executes at multiple sites must either be committed at all the sites, or aborted at all the sites. ­ not acceptable to have a transaction committed at one site and aborted at another · The two-phase commit (2PC) protocol is widely used -- will study this first · The three-phase commit (3PC) protocol is more complicated and more expensive, but avoids some drawbacks of two-phase commit protocol. Database Systems Concepts 18.36 Silberschatz, Korth and Sudarshan c 1997 ' & $ Two-Phase Commit Protocol (2PC) · Assumes fail-stop model ­ failed sites simply stop working, and do not cause any other harm, such as sending incorrect messages to other sites. · Execution of the protocol is initiated by the coordinator after the last step of the transaction has been reached · The protocol involves all the local sites at which the transaction executed · Let T be a transaction initiated at site Si , and let the transaction coordinator at Si be Ci . Database Systems Concepts 18.37 Silberschatz, Korth and Sudarshan c 1997 ' & $ Phase 1: Obtaining a Decision · Coordinator asks all participants to prepare to commit transaction Ti ­ Ci adds the record to the log and forces log to stable storage ­ sends prepare T message to all sites at which T executed · Upon receiving message, transaction manager at site determines if it can commit the transaction ­ if not, add a record to the log and send abort T message to Ci ­ if the transaction can be committed, then: add the record to the log force all log records for T to stable storage send ready T message to Ci Database Systems Concepts 18.38 Silberschatz, Korth and Sudarshan c 1997 ' & $ Phase 2: Recording the Decision · T can be committed if Ci received a ready T message from all the participating sites; otherwise T must be aborted · Coordinator adds a decision record, or , to the log and forces record onto stable storage Once that record reaches stable storage it is irrevocable (even if failures occur) · Coordinator sends a message to each participant informing it of the decision (commit or abort) · Participants take appropriate action locally Database Systems Concepts 18.39 Silberschatz, Korth and Sudarshan c 1997 ' & $ Handling of Failures ­ Site Failure When site Sk recovers, it examines its log to determine the fate of transactions active at the time of the failure · Log contains record: site executes redo(T). · Log contains record: site executes undo(T). · Log contains record: site must consult Ci to determine the fate of T. ­ if T committed, redo(T) ­ if T aborted, undo(T) · The log contains no control records concerning T: implies that Sk failed before responding to the prepare T message from Ci ­ since the failure of Sk precludes the sending of such a response, Ci must abort T ­ Sk must execute undo(T) Database Systems Concepts 18.40 Silberschatz, Korth and Sudarshan c 1997 ' & $ Handling of Failures ­ Coordinator Failure If coordinator fails while the commit protocol for T is executing, then participating sites must decide on T's fate: · If an active site contains a record in its log, then T must be committed. · If an active site contains an record in its log, then T must be aborted. · If some active site does not contain a record in its log, then the failed coordinator Ci cannot have decided to commit T. Can therefore abort T. · If none of the above cases holds, then all active sites must have a record in their logs, but no additional control records (such as or ). In this case active sites must wait for Ci to recover, to find decision. · Blocking problem: active sites may have to wait for failed coordinator to recover. Database Systems Concepts 18.41 Silberschatz, Korth and Sudarshan c 1997 ' & $ Handling of Failures ­ Network Partition · If the coordinator and all its participants remain in one partition, the failure has no effect on the commit protocol. · If the coordinator and its participants belong to several partitions: ­ Sites that are not in the partition containing the coordinator think the coordinator has failed, and execute the protocol to deal with failure of the coordinator. No harm results, but sites may still have to wait for decision from coordinator ­ The coordinator and the sites that are in the same partition as the coordinator think that the sites in the other partition have failed, and follow the usual commit protocol. Again, no harm results Database Systems Concepts 18.42 Silberschatz, Korth and Sudarshan c 1997 ' & $ Recovery and Concurrency Control · In-doubt transactions have a , but neither a , nor an log record. · The recovering site must determine the commit­abort status of such transactions by contacting other sites; this can slow and potentially block recovery. · Recovery algorithms can note lock information in the log. ­ Instead of , write out L = list of locks held by T when the log is written (read locks can be omitted). ­ For every in-doubt transaction T, all the locks noted in the log record are reacquired. · After lock reacquisition, transaction processing can resume; the commit or rollback of in-doubt transactions is performed concurrently with the execution of new transactions. Database Systems Concepts 18.43 Silberschatz, Korth and Sudarshan c 1997 ' & $ Three Phase Commit (3PC) · Assumptions: ­ No network partitioning ­ At any point, at least one site must be up. ­ At most K sites (participants as well as coordinator) can fail · Phase 1: Obtaining Preliminary Decision: Identical to 2PC Phase 1. ­ Every site is ready to commit if instructed to do so ­ Under 2PC each site is obligated to wait for decision from coordinator ­ Under 3PC, knowledge of pre-commit decision can be used to commit despite coordinator failure Database Systems Concepts 18.44 Silberschatz, Korth and Sudarshan c 1997 ' & $ Phase 2: Recording the Preliminary Decision · Coordinator adds a decision record ( or ) in its log and forces record to stable storage · Coordinator sends a message to each participant informing it of the decision · Participant records decision in its log ­ If abort decision reached then participant aborts locally ­ If pre-commit decision reached then participant replies with Database Systems Concepts 18.45 Silberschatz, Korth and Sudarshan c 1997 ' & $ Phase 3: Recording Decision in the Database Executed only if decision in phase 2 was to precommit · Coordinator collects acknowledgments. It sends message to the participants as soon as it receives K acknowledgments. · Coordinator adds the record in its log and forces record to stable storage. · Coordinator sends a message to each participant to . · Participants take appropriate action locally Database Systems Concepts 18.46 Silberschatz, Korth and Sudarshan c 1997 ' & $ Handling Site Failure Site Failure. Upon recovery, a participating site examines its log and does the following: · Log contains record: site executes redo(T). · Log contains record: site executes undo(T). · Log contains record, but no or record: site consults Ci to determine the fate of T. ­ if Ci says T aborted, site executes undo(T) (and writes record) ­ if Ci says T committed, site executes redo(T) (and writes record) ­ if Ci says T precommitted, site resumes the protocol from receipt of precommit T message (thus recording in the log, and sending acknowledge T message sent to coordinator) Database Systems Concepts 18.47 Silberschatz, Korth and Sudarshan c 1997 ' & $ Handling Site Failure (Cont.) · Log contains record, but no or record: site consults Ci to determine the fate of transaction T. ­ if Ci says T aborted, site executes undo(T) ­ if Ci says T committed, site executes redo(T) ­ if Ci says T still in precommit state, site resumes protocol at this point · Log contains no record for a transaction T: site executes undo(T) and writes record. Database Systems Concepts 18.48 Silberschatz, Korth and Sudarshan c 1997 ' & $ Coordinator-Failure Protocol 1. The active participating sites select a new coordinator, Cnew 2. Cnew requests local status of T from each participating site 3. Each participating site, including Cnew, determines the local status of T: · Committed. The log contains a record. · Aborted. The log contains an record. · Ready. The log contains a record but no or record. · Precommitted. The log contains a record but no or record. · Not ready. The log contains neither a nor an record. A site that failed and recovered must ignore any precommit record in its log when determining its status. 4. Each participating site sends its local status to Cnew. Database Systems Concepts 18.49 Silberschatz, Korth and Sudarshan c 1997 ' & $ Coordinator Failure Protocol (Cont.) 5. Cnew decides either to commit or abort T, or to restart the three-phase commit protocol: · Commit state for any one participant commit · Abort state for any one participant abort · Precommit state for any one participant and above 2 cases do not hold A precommit message is sent to those participants in the uncertain state. Protocol is resumed from that point. · Uncertain state at all live participants abort Since at least n - k sites are up, the fact that all participants are in an uncertain state means that the coordinator has not sent a message, implying that no site has committed T. Database Systems Concepts 18.50 Silberschatz, Korth and Sudarshan c 1997 ' & $ Coordinator Selection · Backup coordinators ­ site which maintains enough information locally to assume the role of coordinator if the actual coordinator fails ­ executes the same algorithms and maintains the same internal state information as the actual coordinator ­ allows fast recovery from coordinator failure, but involves overhead during normal processing · Election algorithms ­ used to elect a new coordinator in case of failures ­ Example: Bully Algorithm--applicable to systems where every site can send a message to every other site Database Systems Concepts 18.51 Silberschatz, Korth and Sudarshan c 1997 ' & $ Bully Algorithm · If site Si sends a request that is not answered by the coordinator within a time interval T, assume that the coordinator has failed; Si tries to elect itself as the new coordinator. · Si sends an election message to every site with a higher identification number, Si then waits for any of these processes to answer within T. · If no response within T, assume that all sites with numbers greater than i have failed; Si elects itself the new coordinator. · If answer is received, Si begins time interval T , waiting to receive a message that a site with a higher identification number has been elected. Database Systems Concepts 18.52 Silberschatz, Korth and Sudarshan c 1997 ' & $ Bully Algorithm (Cont.) · If no message is sent within T , assume the site with a higher number has failed; Si restarts the algorithm. · After a failed site recovers, it immediately begins execution of the same algorithm. · If there are no active sites with higher numbers, the recovered site forces all processes with lower numbers to let it become the coordinator site, even if there is a currently active coordinator with a lower number. Database Systems Concepts 18.53 Silberschatz, Korth and Sudarshan c 1997 ' & $ Concurrency Control · Modify concurrency control schemes for use in distributed environment. · We assume that each site participates in the execution of a commit protocol to ensure global transaction atomicity. Database Systems Concepts 18.54 Silberschatz, Korth and Sudarshan c 1997 ' & $ Single-Lock-Manager Approach · System maintains a single lock manager that resides in a single chosen site, say Si . · When a transaction needs to lock a data item, it sends a lock request to Si and lock manager determines whether the lock can be granted immediately ­ If yes, lock manager sends a message to the site which initiated the request ­ If no, request is delayed until it can be granted, at which time a message is sent to the initiating site Database Systems Concepts 18.55 Silberschatz, Korth and Sudarshan c 1997 ' & $ Single-Lock-Manager Approach (Cont.) · The transaction can read the data item from any one of the sites at which a replica of the data item resides. · In the case of a write, all the sites where a replica of the data item resides must be involved in the writing. · Advantages of scheme: ­ Simple implementation ­ Simple deadlock handling · Disadvantages of scheme are: ­ Bottleneck: lock manager site becomes a bottleneck ­ Vulnerability: system is vulnerable to lock manager site failure Database Systems Concepts 18.56 Silberschatz, Korth and Sudarshan c 1997 ' & $ Majority Protocol · Local lock manager at each site administers lock and unlock requests for data items stored at that site. · When a transaction wishes to lock an unreplicated data item Q residing at site Si, a message is sent to Si 's lock manager. · If Q is locked in an incompatible mode, then the request is delayed until it can be granted. · When the lock request can be granted, the lock manager sends a message back to the initiator indicating that the lock request has been granted. · Advantage of simple implementation, however, since lock and unlock requests are no longer made at a single site, deadlock handling is more complex. Database Systems Concepts 18.57 Silberschatz, Korth and Sudarshan c 1997 ' & $ Majority Protocol (Cont.) · In case of replicated data, majority protocol is more complicated to implement than the previous schemes ­ If Q is replicated at n sites, then a lock request message must be sent to more than half of the n sites in which Q is stored. ­ The transaction does not operate on Q until it has obtained a lock on a majority of the replicas of Q. ­ When writing the data item, transaction performs writes on all replicas. · Requires 2(n/ 2 + 1) messages for handling lock requests, and (n/ 2 + 1) messages for handling unlock requests. · Potential for deadlock even with single item -- e.g., each of 3 transactions may have locks on 1/3rd of the replicas of a data item Database Systems Concepts 18.58 Silberschatz, Korth and Sudarshan c 1997 ' & $ Biased Protocol · Local lock manager at each site as in majority protocol, however, requests for shared locks are handled differently than requests for exclusive locks. ­ Shared locks. When a transaction needs to lock data item Q, it simply requests a lock on Q from the lock manager at one site containing a replica of Q. ­ Exclusive locks. When a transaction needs to lock data item Q, it requests a lock on Q from the lock manager at all sites containing a replica of Q. · Advantage -- imposes less overhead on read operations. · Disadvantage -- additional overhead on writes and complexity in handling deadlock. Database Systems Concepts 18.59 Silberschatz, Korth and Sudarshan c 1997 ' & $ Primary Copy · Choose one replica to be the primary copy, which must reside in precisely one site (e.g., primary site of Q). · When a transaction needs to lock a data item Q, it requests a lock at the primary site of Q. · Concurrency control for replicated data handled similarly to unreplicated data--simple implementation. · If the primary site of Q fails, Q is inaccessible even though other sites containing a replica may be accessible. Database Systems Concepts 18.60 Silberschatz, Korth and Sudarshan c 1997 ' & $ Timestamping · Each site generates a unique local timestamp using either a logical counter or the local clock. · Global unique timestamp is obtained by concatenating the unique local timestamp with the unique site identifier. locally-unique globally-unique timestamp site identifier Database Systems Concepts 18.61 Silberschatz, Korth and Sudarshan c 1997 ' & $ Timestamping (Cont.) · A site with a slow clock will assign smaller timestamps "disadvantages" transactions · Define within each site Si a logical clock (LCi ), which generates the unique local timestamp · Require that Si advance its logical clock whenever a transaction Ti with timestamp visits that site and x is greater than the current value of LCi . · In this case, site Si advances its logical clock to the value x + 1. Database Systems Concepts 18.62 Silberschatz, Korth and Sudarshan c 1997 ' & $ Deadlock Handling Consider the following two transactions and history: T1: write(X) T2: write(Y) write(Y) write(X) T1 T2 X-lock on X write(X) X-lock on Y write(Y) wait for X-lock on X wait for X-lock on Y deadlock Database Systems Concepts 18.63 Silberschatz, Korth and Sudarshan c 1997 ' & $ Centralized Approach · A global wait-for graph is constructed and maintained in a single site: the deadlock-detection coordinator. ­ Real graph: Real, but unknown, state of the system. ­ Constructed graph: Approximation generated by the controller during the execution of its algorithm. · The global wait-for graph can be constructed when: ­ a new edge is inserted in or removed from one of the local wait-for graphs. ­ a number of changes have occurred in a local wait-for graph. ­ the coordinator needs to invoke cycle-detection. · If the coordinator finds a cycle, it selects a victim and notifies all sites. The sites roll back the victim transaction. Database Systems Concepts 18.64 Silberschatz, Korth and Sudarshan c 1997 ' & $ Centralized Approach (Cont.) · Unnecessary rollbacks may result when a deadlock has indeed occurred and a victim has been picked, and meanwhile one of the transactions was aborted for reasons unrelated to the deadlock. · Unnecessary rollbacks can result from false cycles in the global wait-for graph; however, likelihood of false cycles is low. Database Systems Concepts 18.65 Silberschatz, Korth and Sudarshan c 1997 ' & $ False Cycles T1 T2 T2 T1 T3 S2 T1 T3 coordinator S1 Database Systems Concepts 18.66 Silberschatz, Korth and Sudarshan c 1997 ' & $ False Cycles (Cont.) · Suppose that starting from state shown in figure, 1. T2 releases resources at S1 (resulting in a message remove T1 T2 message from the Transaction Manager at site S1 to the coordinator), and then 2. T2 requests a resource held by T3, at site S2 (resulting in a message insert T2 T3 from S2 to the coordinator) · Suppose further that the insert message reaches before the delete message. The coordinator would then find a false cycle T1 T2 T3 T1 · The false cycle above never existed in reality. · False cycles cannot occur if two-phase locking is used Database Systems Concepts 18.67 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach · Each site has local wait-for graph; system combines information in these graphs to detect deadlock · Local Wait-for Graphs Site 1 Site 2 Site 3 T1 T2 T3 T3 T4 T5 T5 T1 · Global Wait-for Graph T1 T2 T3 T4 T5 | Database Systems Concepts 18.68 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach (Cont.) · System model: a transaction runs at a single site, and makes requests to other sites for accessing non-local data. · Each site maintains its own local wait-for graph in the normal fashion: there is an edge Ti Tj if Ti is waiting on a lock held by Tj (note: Ti and Tj may be non-local). · Additionally, arc Ti Tex exists in the graph at site Sk if (a) Ti is executing at site Sk , and is waiting for a reply to a request made on another site, or (b) Ti is non-local to site Sk , and a lock has been granted to Ti at Sk . · Similarly, arc Tex Ti exists in the graph at site Sk if (a) Ti is non-local to site Sk , and is waiting on a lock for data at site Sk , or (b) Ti is local to site Sk , and has accessed data from an external site. Database Systems Concepts 18.69 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach (Cont.) · Centralized Deadlock Detection -- all graph edges sent to central deadlock detector · Distributed Deadlock Detection -- "path pushing" algorithm · Path pushing initiated when a site detects a local cycle involving Tex , which indicates possibility of a deadlock. · Suppose cycle at site Si is Tex Ti Tj . . . Tn Tex and Tn is waiting for some transaction at site Sj . Then Si passes on information about the cycle to Sj · Optimization: Si passes on information only if i > n. · Sj updates it graph with new information and if it finds a cycle it repeats above process. Database Systems Concepts 18.70 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach: Example Site 1 EX(3) T1 T2 T3 EX(2) Site 2 EX(1) T3 T4 T5 EX(3) Site 3 EX(2) T5 T1 EX(1) EX(i): Indicates Tex , plus wait is on/by a transaction at Site i Database Systems Concepts 18.71 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach Example (Cont.) · Site passes wait-for information along path in graph: ­ Let EX(j) Ti ... Tn EX(k) be a path in the local wait-for graph at Site m ­ Site m "pushes" the path information to site k if i > n · Example: ­ Site 1 does not pass information : 1 < 3 ­ Site 2 does not pass information : 3 < 5 ­ Site 3 passes (T5, T1) to Site 1 because: 5 > 1 T1 is waiting for a data item at site 1 Database Systems Concepts 18.72 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach (Cont.) · After the path EX(2) T5 T1 EX(1) has been pushed to Site 1 we have: Site 1 EX(2) T5 T1 T2 T3 EX(2) Site 2 EX(1) T3 T4 T5 EX(3) Site 3 EX(2) T5 T1 EX(1) Database Systems Concepts 18.73 Silberschatz, Korth and Sudarshan c 1997 ' & $ Fully Distributed Approach (Cont.) · After the push, only Site 1 has new edges. Site 1 passes (T5, T1, T2, T3) to site 2 since 5 > 3 and T3 is waiting for a data item at site 2 · The new state of the local wait-for graph: Site 1 EX(2) T5 T1 T2 T3 EX(2) Site 2 T5 T1 T2 T3 T4 | Deadlock Detected Site 3 EX(2) T5 T1 EX(1) Database Systems Concepts 18.74 Silberschatz, Korth and Sudarshan c 1997 ' & $ Multidatabase Systems · Software layer on top of existing database systems required to manipulate information in heterogeneous database · Data models may differ (hierarchical, relational, etc.) · Transaction commit protocols may be incompatible · Concurrency control may be based on different techniques (locking, timestamping, etc.) · System-level details almost certainly are totally incompatible Database Systems Concepts 18.75 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advantages · Preservation of investment in existing ­ hardware ­ systems software ­ applications · Local autonomy and administrative control · Allows use of special-purpose DBMSs · Step towards a unified homogeneous DBMS Database Systems Concepts 18.76 Silberschatz, Korth and Sudarshan c 1997 ' & $ Unified View of Data · Agreement on a common data model · Agreement on a common conceptual schema · Agreement on a single representation of shared data (that may be stored in multiple DBMSs) · Agreement on units of measure · Willingness to accept limited function in global transactions Database Systems Concepts 18.77 Silberschatz, Korth and Sudarshan c 1997 ' & $ Transaction Management · Local transactions are executed by each local DBMS, outside of the MDBS system control. · Global transactions are executed under MDBS control. · Local autonomy--local DBMSs cannot communicate directly to synchronize global transaction execution and the MDBS has no control over local transaction execution. ­ local concurrency control scheme needed to ensure that DBMS's schedule is serializable ­ in case of locking, DBMS must be able to guard against local deadlocks ­ need additional mechanisms to ensure global serializability Database Systems Concepts 18.78 Silberschatz, Korth and Sudarshan c 1997 ' & $ Two-Level Serializability · DBMS ensures local serializability among its local transactions, including those that are part of a global transaction. · The MDBS ensures serializability among global transactions alone -- ignoring the orderings induced by local transactions. · 2LSR does not ensure global serializability, however, it can fulfill requirements for strong correctness: 1. Preserve consistency as specified by a given set of constraints 2. Guarantee that the set of data items read by each transaction is consistent · Global-read protocol: Global transactions can read, but not update, local data items; local transactions do not have access to global data. There are no consistency constraints between local and global data items. Database Systems Concepts 18.79 Silberschatz, Korth and Sudarshan c 1997 ' & $ Two-Level Serializability (Cont.) · Local-read protocol: Local transactions have read access to global data; disallows all access to local data by global transactions. ­ A transaction has a value dependency if the value that it writes to a data item at one site depends on a value that it read for a data item on another site. ­ For strong correctness: No transaction may have a value dependency. · Global-read­write/local-read protocol: Local transactions have read access to global data; global transactions may read and write all data; ­ No consistency constraints between local and global data items. ­ No transaction may have a value dependency. Database Systems Concepts 18.80 Silberschatz, Korth and Sudarshan c 1997 ' & $ Global Serializability · Global 2PL--each local site uses a strict 2PL (locks are released at the end); locks set as a result of a global transaction are released only when that transaction reaches the end. · Even if no information is available concerning the structure of the various local concurrency control schemes, a very restrictive protocol that ensures serializability is available. ­ Transaction-graph: a graph with vertices being global transaction names and site names. An undirected edge (Ti , Sk ) exists if Ti is active at site Sk . ­ Global serializability is assured if transaction-graph contains no undirected cycles. Database Systems Concepts 18.81 Silberschatz, Korth and Sudarshan c 1997 ' & $ Ensuring Global Serializability · Each site Si has a special data item, called a ticket · Every transaction Tj that runs at site Sk writes to the ticket at site Si · Ensures global transactions are serialized at each site, regardless of local concurrency control method, so long as the method guarantees local serializability · Global transaction manager decides serial ordering of global transactions by controlling order in which tickets are accessed · However, above protocol results in low concurrency between global transactions Database Systems Concepts 18.82 Silberschatz, Korth and Sudarshan c 1997