C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986. 14  Home/Search   Document Not in Database   Summary   Related Articles   Check

....IS IX S SIX IS IX S SIX X X Table 6.1: Lock mode compatibility matrix for granular locks. The purpose of the various lock modes are shown alongside. Serializability Concepts and the Phantom Problem Transactions, locking and serializability concepts are well documented in the literature [112, 113, 55]. The phantom problem is defined as follows (from the ANSI ISO SQL 92 specifications [93, 7] Transaction T1 reads a set of data items satisfying some search condition . Transaction T2 then creates data items that satisfy T1 s search condition and commits. If T1 then repeats its scan with the ....

C. Papadimitriou. The theory of database concurrency control. Computer Science Press, 1986.

....the key concepts here is the notion of serializability, i.e. that the outcome of concurrently executed transactions is equivalent to a strictly serial execution of the transactions. Most of the protocols that guarantee serializability already found their way into textbooks more than a decade ago [2, 10, 14]. During the last decade some researchers have concentrated on defining notions weaker than serializability and developed protocols that allow a more liberal cooperation between users. For a survey on cooperating transactions and synchronization in general see [15] However, we believe that ....

....therefore talk about nodes only. 3 Protocols In this section we introduce the core protocols for synchronizing structure traversals and modifications of XML documents. We also give some more details necessary for the explanations. Generally speaking, our protocols are based on two phase locking [2, 14]. It is important to note that all core protocols require that document access starts at the root node and traverses documents top down. This requirement is relaxed in Section 5. 3.1 Lock Modes In standard two phase locking protocols for synchronizing read and write operations, we have two kinds ....

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986. 14

....Figure 1: Two seemingly correct scenarios, updating fuel amounts the executions are partially ordered and we infer distributed implementations. 2 Sample MSC Inference We motivate inference of missing scenarios using an example related to serializability in database transactions (see, e.g. [23]) Consider the following standard example, described in the setting of a nuclear power plant. Two clients, P 1 and P 2 , seek to perform remote updates on data used in the control of a nuclear power plant. In this database the variable UR controls the amount of Uranium fuel in the daily supply at ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

....for ordered time and not for the generalized metrics of [4] leaving this as a loose end. In between [4] and [1] we introduced the notion of higher dimensional automaton (HDA) 6] as an algebraic topological form of automata theory supporting Papadimitriou s geometric view of concurrency control [7] in terms of higherdimensional state spaces, in which mutual exclusion takes the form of a hole. At that time we were unable to answer Boris Trakhtenbrot s question after our POPL talk as to how HDAs were related to event spaces, leaving another loose end which we only recently tied up using ....

Papadimitriou, C.: The Theory of Database Concurrency Control. Computer Science Press (1986)

....of XML document collections. The development of synchronization protocols to isolate different applications has a long standing and successful history in the database community. Most of the protocols that guarantee serializability already found their way into textbooks more than a decade ago [2, 9, 14]. During the last decade some researchers concentrated on defining notions weaker than serializability and developed protocols that allow cooperation between users. For a recent survey on cooperating transactions and synchronization in general see [15] Although cooperation will play a major role ....

....to synchronize structure traversals and modifications of XML documents. We start by introducing several protocols based on two phase locking. Next, we introduce a core protocol based on timestamp ordering. Both protocols are well known and already found their way into several textbooks, e.g. [2, 14]. A novel feature of our variant of the timestamp ordering protocol is that 4 T M T M (a) TL TR TA TZ ML MR MA MZ TL TR TA TZ ML MR MA MZ (b) ....

[Article contains additional citation context not shown here]

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

....global database is in a correct (consistent) state, if and only if all its component databases are in correct (consistent) states. 2.2 Effects of Local Autonomies and Indirect Conflicts In the literature of concurrency control, serializability is usually used as the correctness criterion. [2, 3] The aim is to keep the execution serializable. However, the goal is difficult to achieve in a multidatabase environment. For example, let G 11 , G 12 be two subtransactions of a global transaction G 1 , and G 21 , G 22 be two subtransactions of another global transaction G 2 , where G ij (i; j = ....

C. Papadimitriou, The Theory of Database Concurrency Control, Computer Science Press, 1986.

....with W i [x i ] When T i reads an item, it reads the value of a version written by some transaction T j , notated like: R i [x j ] Each element from I is read or written once by a given transaction. All read actions are executed before the write actions. This often justified simplification [9] renders the algorithms and their accompanying proofs much simpler. An example of the execution of transaction T i is: T i = R i [x j ]R i [y k ]W i [x i ]W i [z i ] The values of j and k depend on the history of the data items. A transaction T i has a read set RS i and a write set WS i defined ....

Papadimitriou, C.H. The theory of Database Concurrency Control. Computer Science Press, 1986.

....paradigm [6; 12] is inherently solved by this approach. 2 The formal model This section develops the schedule based formal model for dynamic actions. For a complete description, the reader is referred to [7] The notion of schedules is the basis for the well developed theory of concurrency [11]. Hadzilacos [4; 5] introduces a Herbrand semantics [3] for schedules, that allows to analyse reliability properties of schedules. However, the achievement of [5] is to formalise conditions how to commit concurrent, isolated computations. In the following, the approach of [5] is extended to ....

C. H. Papadimitriou. The Theory of Database Concurrency Control. Rockville, Computer Science Press, 1986

....performance, the e ect must be the same as if transactions were executed sequentially. Transaction throughput is a crucial issue for all databases. The common approach in practice considers con ict serializable schedules, where con icts correspond to read and write operations on database objects [8, 20]. No matter which granularity is taken for these objects pages, records or even relations occur in practice this approach rules out acceptable, but formally not serializable schedules. In order to increase the rate of concurrency multi level transactions (as a special form of nested ....

....of a global level, a local logical object level, a local level of physical objects and a page level. This is the view adopted in the DOMOCC project currently under investigation at Clausthal. The general approach to concurrency control is the use of locking protocols, especially two phase locking [20]. It will be shown how to generalize lock protocols to multi level transactions. This will ll Section 3. The major problems with this approach are transaction throughput and the possibility of deadlocks due to transactions waiting for each other to release locks. There are several algorithms for ....

C. Papadimitriou. The Theory of Database Concurrency Control . Computer Science Press, 1986.

.... reserve(A) modify(p1) write(A) reserve(A) modify(p2) write(A) reserve(B) modify(p3) write(B) reserve(B) modify(p4) write(B) Time Figure 2: Serializable schedule or concurrently with other transactions in a multi programmed system [Bernstein et al. 87; Papadimitriou 86] Let us follow up on our previous example to demonstrate the transaction concept. John and Mary are now assigned the task of fixing two bugs that were suspected to be in modules A and B. The first bug is caused by an error in procedure p1 in module A, which is called by procedure p3 in module ....

Papadimitriou, C. The Theory of Database Concurrency Control. Computer Science Press, Rochville, MD, 1986. 77

....at one extreme, to application dependent, special purpose compensating transactions, at the other extreme. We return to study the characteristics of compensating transactions in more depth after we introduce the model in the next section. 3 A Transaction Model In the classical transaction model [Pap86, BHG87] transactions are viewed as sequences of read and write operations that map consistent database states to consistent states when executed in isolation. The correctness criterion of this model is called serializability. A concurrent execution of a set of transactions is represented as 4 an ....

....not commute. Part of the orderings implied by the total order in which operations are composed to form a history are arbitrary, since only conflicting operations must be totally ordered. In essence, our equivalence notion (when restricted to database state) is similar to final state equivalence [Pap86] However, in what follows, we shall need to equate histories that are not necessarily over the same set of transactions, which is in contrast to final state equivalence (and actually to all familiar equivalence notions) A projection of a history X on an entity e is is a subsequence of X, that ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

....to a version as an alternative. Thus, the alternatives are expected to be managed by the applications themselves. When not specified 3 explicitly, the last written alternative will be regarded as being the one to which a reference is being made. Note that in the standard models (e.g. see [BHG87, Pap86] only one alternative of an entity (i.e. the standard alternative see [KS88, Pap86] is accessible. Each entity is permitted to have a set of application dependent attributes. For example, an attribute timestamp(d) for an entity d may be regarded as a temporal attribute for the entity d. ....

....themselves. When not specified 3 explicitly, the last written alternative will be regarded as being the one to which a reference is being made. Note that in the standard models (e.g. see [BHG87, Pap86] only one alternative of an entity (i.e. the standard alternative see [KS88, Pap86] is accessible. Each entity is permitted to have a set of application dependent attributes. For example, an attribute timestamp(d) for an entity d may be regarded as a temporal attribute for the entity d. Note that the semantics associated with timestamp(d) are entirely application dependent in ....

[Article contains additional citation context not shown here]

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

....executing transactions in a serial order. Since these protocols and implementations, in general, yield poor performance, a significant amount of research has been done to devise methods to obtain better performance. The two most common techniques for achieving this are pipelining and concurrency [1, 2, 6, 7, 8, 9, 10, 11]. Pipelining is a method for overlapping the execution of multiple operations of a process, whereas concurrency is a method for overlapping the execution of multiple operations of different processes (a process is an execution of a sequential program, namely, an execution of a sequential operation ....

....operation R(object) We assume that the system stalls before issuing a write operation until all reads that write is dependent are performed. Hence, for any write operation a that is dependent on a read operation b, b p a. 15 3. 6 Interpretation We borrow the notion of interpretation from [7]. The interpretation of a schedule (IB i ) is specified by the program of process P i from which the schedule is originated. An interpretation of a schedule is a pair I i = D; F ) where D = fD x ; D y ; g is a set of domains, one for each entity in E; each domain is a set of values for the ....

[Article contains additional citation context not shown here]

C. Papadimitriou, The Theory of Database Concurrency Control, Computer Science Press, 1986.

....executions. The approach we adopt in this paper enables the semantics of transactions to be exploited, and is similar to the one used in [GM83, FO89] However, our mechanism for specifying interleavings is based 1 In this paper, we limit ourselves to conflict serializability (CSR) Pap86] which we shall refer to, in the remainder of the paper, as serializability. 2 on regular expressions over transaction types, and many of the interleavings that can be specified using our approach cannot be specified using the approaches in [GM83, FO89] The proposed mechanism assumes that the ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

....known theorems lead to e#cient algorithms for checking the atomicity of a run in certain special cases. To our knowledge, the atomicity problem has not been studied before from a complexity theory point of view (in contrast to the serializability problem for transactions on a database, see e.g. [18]) Detailed proofs are not given, since what is proved in the next section, which is the main contribution of this paper, covers as special cases all the results of this section. 2.1 Run on a Single Variable In this subsection we assume that c = 1, i.e. we assume that the run is on a single ....

....The complexity of this algorithm is the same as the complexity of the algorithm for the decision problem. Discussion It is interesting to notice that closely related serializability problems in database concurrency control have been shown to be NP complete. For example, it is proved in [18] that view serializability is an NP complete problem (a view serialization is a serialization of a history of transactions on a database in such a way as the final state of the database is the same for the serial and the given history and, moreover, the the read steps of all transactions return ....

C. Papadimitriou (1986): The Theory of Database Concurrency Control, Computer Science Press.

....The next sub section describes a standard concurrency control algorithm from databases and identifies the problems in applying it to knowledge bases directly. 1. 5 Problem in Using Conventional Concurrency Control Algorithms There is a vast body of literature on concurrency control algorithms (Papadimitriou 1986; Bernstein, Hadzilacos and Goodman 1987; Gray and Reuter 1993) There are three broad classes of such algorithms: locking, timestamps and serialization graphs. For each of these classes, there are variations based on multiple versions and certifiers. Lockingbased algorithms have been most ....

Papadimitriou, C. 1986. The Theory of Database Concurrency Control. Computer Science Press, Rockville, MD.

....reconstruct an older state of a KB if necessary. 7.1.7 Correctness The conflict matrix defined in Table 2 should have the property that merging conflict free updates to the public copy of the KB always leads to correct KB states. Correctness is commonly captured by the notion of serializability (Papadimitriou 1986). Serializability can be informally defined as follows. Let a transaction be the set of operations executed by a user. An interleaved execution of a set of transactions is serializable if it is equivalent to some serial execution of the same set of transactions in the sense that it 34 leaves the ....

Papadimitriou, C. 1986. The Theory of Database Concurrency Control. Computer Science Press, Rockville, MD.

.... Autonomy: It is not practically feasible to modify the underlying local DBMS software to facilitate integration. Database consistency is traditionally ensured by requiring that the concurrent execution of transactions be serializable #that is, equivalent to a non concurrent execution##Papadimitriou 1986#. The problem of ensuring global serializabilityin an MDBS environment has been studied extensively #Breitbart and Silberschatz 1988; Breitbart et al. 1990; Georgakopoulos et al. 1991; Batra et al. 1992; Du et al. 1989; Mehrotra et al. 1992; Pu 1988; Raz 1992#. A necessary condition for # 3 ....

....constraints, denoted by IC, in a database distinguish inconsistent database states from consistent ones. One way to formalize the notion of integrity constraints is to consider them as a subset of all the possible database states, and a database state is consistent if it belongs to that subset #Papadimitriou 1986#. An equivalent formulation is to consider integrity constraints as a conjunction of #rst order logic formulae over a language consisting of the following: Numerical and string constants #e.g. 5, 100, Jim #, Functions over numeric and string constants #e.g. max#, Comparison ....

Papadimitriou, C. 1986. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland.

....of database consistency in terms of the preservation of integrity constraints. We then develop our transaction model to help us in reasoning about non serializable executions. Finally, we discuss our notion of correctness of a schedule. 2. 1 Database Consistency In the standard transaction model [13], a consistent database state is implicitly defined by assuming that each transaction, when executed in isolation, maps a consistent database state to another consistent database state. Correctness in case of concurrent execution is defined in terms of serializability. In order to develop a theory ....

....0 1 6= v 0 2 . 5 Integrity constraints in a database, denoted by IC, distinguish inconsistent database states from consistent ones. Traditionally, integrity constraints are defined as a subset of all the possible database states, and a database state is consistent if it belongs to that subset [13]. In our model, integrity constraints are quantifier free first order formulae over the language consisting of: ffl Numerical and string constants (e.g. 5, 100, Jim ) ffl Functions over numeric and string constants (e.g. max) ffl Comparison operators (e.g. and ffl Set of ....

[Article contains additional citation context not shown here]

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986. 26

....Their algorithms observe the sequential stream of operations at the data structure, and check to see if this stream is legal; in contrast, our main task is to determine if there exists a legal sequential stream corresponding to the collection of parallel streams that we observe. Papadimitriou [Pap86] proved that testing the serializability of database transactions is NP complete. Gibbons and Korach [GK97] studied testing a shared memory for sequential consistency or linearizability under a range of scenarios. They did not explicitly consider testing executions of data structures. Other work ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

..... 1 A set P 1 with a partial order OE P 1 on its elements is a restriction of a set P 2 with a partial order OE P 2 on its elements if P 1 P 2 , and for all e 1 ; e 2 2 P1 , e 1 OE P 1 e 2 if and only if e1 OE P 2 e 2 . 2 In this paper, we limit ourselves to conflict serializability (CSR) Pap86] which we shall refer to, in the remainder of the paper, as serializability. 2.2 Serialization Functions In order to develop our idea, we need to first introduce the notion of serialization functions, which is similar to the notion of serialization events [ED90] Let k be the set of all ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

....will also be correct. Here what we mean by correct execution is an execution that preserves database consistency. However in a concurrent system, due to the interleaving of accesses, an execution that would be correct if the program was running alone, can result in an inconsistent database [Pap86, BHG87] This is due to the non controlled interference of operations over the shared data. This gives rise to two requirements: ffl We want the database system to take advantage of multiprogrammed execution of application programs in order to have efficient system utilization. ffl We do not ....

....we will adopt for nested transactions interprets 8 the notion of state transformation and return values. In this latter model we have two formal definition of commutativity one that is state dependent and the other state independent. Transactions can be described as sequences of operations [Pap86] or as a partial order of operations [BHG87] The advantage of modeling transactions as a partial order is that the transaction need not specify the ordering of every two operations that appear in it. Beyond being a sequence or a partial order of operations some conditions must be satisfied. ....

[Article contains additional citation context not shown here]

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

....[6, 3] In our setting, only positive examples are given, but the executions are partially ordered and we infer distributed implementations. 2 SAMPLE MSC INFERENCE We motivate inference of missing scenarios using an example related to serializability in database transactions (see, e.g. [18]) Consider the following standard example, described in the setting of a nuclear power plant. Two clients, P 1 and P 2 , seek to perform remote updates on data used in the control of a nuclear power plant. In this database the variable UR controls the amount of Uranium fuel in the daily supply ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

....out in detail to illustrate the application of the ELL supervisor. We have made some simplifying assumptions so that we are able to focus on the main issues. In the following, we introduce some terminologies for concurrency control of transaction execution in DBMS; interested readers may refer to [21, 22] for details. A transaction is defined to be a sequence of read and write operations on the data items of the database. The additional operation, called commit operation, is used to signify successful termination of the transaction. A transaction that is not committed is called active. Let X ....

....to this implementation issue is that instead of taking set intersection to compute z [Kns] N , it is possible to perform and implement a legality test on z [P ns] N to generate z [Kns] N . For instance, in the case of DBMS, Serialization Graph Testing (SGT) [22, 21] could be used to determine which schedules are serializable. 3. The non blocking feature often results in restrictive control in the closed loop behavior; thus, the generated controlled behavior, although non blocking, may be restrictive. As noted by Chen and Lafortune [3] it is sometime better ....

[Article contains additional citation context not shown here]

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, MD, 1986.

....OF TEXAS AT AUSTIN austin, texas 78712 On Correctness of Non serializable Executions 3 Sharad Mehrotra Rajeev Rastogi Henry F. Korth Avi Silberschatz Department of Computer Sciences University of Texas at Austin Austin, TX 78712 1188 USA 1 Introduction In the standard transaction model [8], a consistent database state is implicitly defined by assuming that each transaction, when executed in isolation, maps a consistent database state to another consistent database state. In the case of concurrent transaction executions, database consistency is ensured by requiring that the schedule ....

....maps a consistent database state to another consistent database state. In the case of concurrent transaction executions, database consistency is ensured by requiring that the schedule resulting from the concurrent executions of transactions be serializable; that is, equivalent to a serial schedule [8]. Since each transaction, when executed alone, is assumed to preserve database consistency, a serializable execution preserves database consistency. This approach has the advantages of simplicity since it does not require the users to state explicitly what constitutes a consistent database state. ....

[Article contains additional citation context not shown here]

C. Papadimitriou, "The Theory of Database Concurrency Control", Computer Science Press, 1986. 12

....conflict graph and the distributed dependencies graph. Based on these graphs, the scheduler searches the best schedule among the feasible 2 ones, that minimizes a virtual function of the management time. The distributed conflict graph is meaningful in the serializability theory [CP85, Pap86, BHN87, GR94] Two operations are in conflict if they both operate on the same data item, one of them is a Write operation, and they are issued by different transactions. The conflict graph has the transactions as its nodes and it contains an arc (A,B) that links transaction A to that B, whenever ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

....In Figure 6, we found that optimistic performs better than locking for very low arrival rates. Details of this study can be found in [4] 16 4 More Ideas for Increasing Concurrency Multidimensional Time Stamps There are several variations of timestamp ordering. For example, multiple versions [25] of item values have been used to increase the degree of concurrency. The conventional time stamp ordering tends to prematurely determine the serializability order, which may not fit in with the subsequent history, forcing some transactions to abort. The multidimensional time stamp protocol [21] ....

.... of two phase locking were discussed in [13] The ideas of time stamps were introduced by [29] The optimistic approach was proposed by [19] The classes of serializability and the formalism for concurrency control was presented in [24] Several books that detail these subjects have been published [6, 25, 2] in addition to survey papers [1, 5] The performance evaluation was studied in [14] In most commercial systems, the most popular mechanism for concurrency control is two phase locking [17] The ideas of adaptable concurrency control were published in [11] and were implemented in Raid system ....

C. H. Papadimitriou, The Theory of Database Concurrency Control, Computer Science Press, 1986.

....with the minimum of effort made by the application programmer. We consider both the problem of synchronising operations invoked by transactions on objects and the problem of synchronising transactions. The execution of transactions should satisfy the correctness criterion of serialisability [20]. An important aspect of flexibility is the provision of more than one protocol. The model can realise a wide range of protocols including the three main categories of concurrency control protocols, as well as hybrid protocols that combine 3 characteristics of the former. The implementation of ....

....one of delay , accept or abort transaction with some preconditions that represent the conflict to be resolved. It is explained in more detail in Section 4. 3. 1 THE NOTION OF CONFLICT Concurrency control protocols attain serialisability by considering conflicts between pairs of operations [3, 7, 20]. Conflicts may arise between operations invoked on the same object only. When a conflict is found actions are taken to resolve it. Informally, two operations conflict if the observations of the one are no longer valid because of the other, or vice versa. In the conventional database protocols two ....

C. Papadimitriou, "The theory of database concurrency control", Computer Science Press, (1986)

....a pair (a e) where a is an action (one of INSERT, DELETE, READ, WRITE) and e is an entity, which is a node or an edge. Using INSERT and DELETE operations, transactions can modify the graph. The transactions can acquire two kinds of locks shared and exclusive, which have their usual semantics [4, 10, 19]. We first define some properties of directed graphs that are necessary for specifying our algorithm. A root of a directed graph is a node that does not have any predecessors. A directed graph is rooted if it has a unique root and there is a path from the root to every other node in the graph. A ....

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, MD, 1986.

....the overall execution is serializable, is called concurrency control algorithm or policy. The next sub section describes some concurrency control algorithms from databases. 2. 3 CONCURRENCY CONTROL ALGORITHMS FROM DATABASES There is a vast body of literature on concurrency control algorithms (Papadimitriou 1986; Bernstein, Hadzilacos and Goodman 1987) There are three broad classes of such algorithms: locking, timestamps and serialization graphs. For each of these classes, there are variations based on multiple versions and optimistic methods. Locking based algorithms have been most successful in ....

....base state they are executed. A partial schedule S of a transaction system is a legal and a proper schedule of any prefixes of the transactions of . In our model, the correctness of a schedule is equivalent to serializability. Serializability of a schedule can be easily decided as follows (Papadimitriou 1986). Let e denote the set of nodes in an entity e (recall that an entity is a node or an edge) Construct a directed graph D(S) by associating a node v i with each transaction T i and including an arc (v i ; v j ) if in schedule S, T i acts on an entity e x before T j does on entity e y and e x e ....

Papadimitriou, C. 1986. The Theory of Database Concurrency Control. Computer Science Press, Rockville, MD.

....between groupware and multiuser databases. Due to this difference it is usually not appropriate to use a conservative database approach to implement groupware. The reason for this is further explained in the next section. In database systems concepts of transaction processing (like locking) Papa 86] are used in order to keep the data at each site consistent. These methods are usually not applicable in groupware systems for several reasons. When locking is used the size of the records is comparably big, which has the effect that other users can not access the data they would like to. Due to ....

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Md, 1986.

....EXECUTION OF RULES In this section, we discuss the Concurrent Execution Process (CEP) in detail. The traditional definition for the correctness of concurrently executing transactions is that the execution must be serializable and equivalent to some serial execution of these transactions [1, 12, 30]. Informally, if two concurrently executing transactions T i and T j update some set of relations R 1 , R 2 , R n , then if T i reads a relation (tuple or page) of some R k which is subsequently updated by T j , or if T i updates a page of R k which is subsequently read or updated by T j , ....

....updated by T j , or if T i updates a page of R k which is subsequently read or updated by T j , then serializability requires that the commit order of these two transactions cannot be such that the execution of T j precedes the execution of T i . This is also called conflict serializability [30]. This commit order then dictates the order in which the transactions actually update the database relations. In a rule, the query corresponding to the condition must be satisfied by the database, following which the actions are executed. One of the key differences from traditional DBMS is that ....

Papadimitriou, C. The Theory of Database Concurrency Control. Computer Science Press, 1986.

....algorithms such as two phase locking, timestamp ordering and serialization graph testing [BHG87] see Section 2.2. 4) Two other forms of serializability, view serializability and final state serializability, exist but are only of theoretical interest because they are known to be NP complete [Pap86] View serializable histories are defined using a definition of view equivalence. Definition 2.2.10 Two histories, H and H 0 , are view equivalent if 1. they are over the same set of transactions and have the same operations, 2. for any T i ; T j such that a i ; a j = 2 H (hence a i ; a j = 2 ....

....final state serializability depend only on the completed state of the transactions, and not any intermediate transaction states, it cannot be used by a dynamic scheduler. The problem of recognizing the two serializability methods is NP complete and likely computationally infeasible to implement [Pap86] 2.2.3 Combining Reliability and Concurrency Serializability guarantees correct concurrent operation orderings when failures do not occur. If failures occur, the schedules produced must be serializable and at least recoverable to guarantee correctness. To coordinate multiple write operations to ....

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

No context found.

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986. 14

No context found.

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

No context found.

C. Papadimitriou, "The Theory of Database Concurrency Control, Computer Science Press, 1986.

No context found.

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

No context found.

C. Papadimitriou. The theory of database concurrency control. Computer Science Press, 1986.

No context found.

C. Papadimitriou. The theory of database concurrency control. Computer Science Press, 1986.

No context found.

C. H. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, 1986.

No context found.

C. Papadimitriou. The theory of database concurrency control. Computer Science Press, 1986.

No context found.

C. Papadimitriou. The Theory of Database Concurrency Control. Principles of Computer Science. Computer Science Press, Rockville, Maryland, 1986.

No context found.

C. Papadimitriou, The Theory of Database Concurrency Control, Computer Science Press, 1986.

No context found.

C.H. Papadimitriou, The theory of database concurrency control. Computer Science Press, Rockville, 1986.

No context found.

Papadimitriou, C. The Theory of Database Concurrency Control, Computer Science Press, 1986.

No context found.

Papadimitriou, C. (1986). The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland.

No context found.

C. Papadimitriou. The Theory of Database Concurrency Control. Computer Science Press, Rockville, Maryland, 1986.

No context found.

PAPA86 Papadimitriou, C. H., The Theory of Database Concurrency Control, Computer Science Press, ISBN 0-88175-027-1, 1986.

No context found.

C. Papadimitriou, The Theory of Database Concurrency Control, Computer Science Press, 1986.

No context found.

PAPA86 Papadimitriou, C. H., The Theory of Database Concurrency Control, Computer Science Press, ISBN 0-88175-027-1, 1986.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC