Update Transactions Research Articles

A formal characterization of epsilon serializability

Epsilon serializability (ESR) is a generalization of classic serializability (SR). In this paper, we provide a precise characterization of ESR when queries that may view inconsistent data run concurrently with consistent update transactions. Our first goal is to understand the behavior of queries in the presence of conflicts and to show how ESR in fact is a generalization of SR. So, using the ACTA framework, we formally express the intertransaction conflicts that are recognized by ESR and through that define ESR, analogous to the manner in which conflict-based serializability is defined. Secondly, expressions are derived for the amount of inconsistency (in a data item) viewed by a query and its effects on the results of a query. These inconsistencies arise from concurrent updates allowed by ESR. Thirdly, in order to maintain the inconsistencies within bounds associated with each query, the expressions are used to determine the preconditions that operations have to satisfy. The results of a query, and the errors in it, depend on what a query does with the (possibly inconsistent) data viewed by it. One of the important byproducts of this work is the identification of different types of queries which lend themselves to an analysis of the effects of data inconsistency on the results of the query.

A rule-based update language for complex objects with identity

This paper deals with the definition of a rule-based update language for complex object databases. Several update operators are introduced, which correspond to basic manipulations of data entities with object identity. These operators can be used in rule bodies for specifying complex update transactions. The syntax and the semantics of the language is given in the framework of CO, a simple data model for complex objects with identity. An interpreter of the language is defined and its correctness is discussed. Finally, the issue of maintaining database consistency with respect to several kinds of integrity constraints is illustrated.

Dependency preservation in semantic databases

A simple semantic or object-based data model is considered, which includes objects and object identifiers, classes and class hierarchies, attributes ranging over atomic values. Transactions are composed from five, basic operators manipulating objects. Preservation of functional and acyclic inclusion dependencies by transactions is studied in such a context of semantic databases and update transactions. It is shown to be decidable whether a given transaction preserves a given set of functional dependencies, or acyclic inclusion dependencies, or both functional and acyclic inclusion dependencies. Time complexity (with respect to the sizes of transactions and database schemas) for testing preservation is also discussed. It turns out that the problem is co-NP-complete in the simplest cases where there is only one nontrivial dependency and transactions consist of only creations and deletions of objects. It implies that the problem is at least co-NP-hard in general.

An entity-based approach to query processing in relational databases: Part II: Entity query construction and updating

A declarative entity-based query language called the E-TRC for relational databases is presented and the construction and evaluation process of RA queries for its expressions is developed. This extends the value-oriented entity representation and the entity-based window generator for entity type projections of Part I. It is also shown how entity-level update transactions can be interpreted in a logical and consistent way and targeted to suitable relations while maintaining database integrity. The query construction and evaluation process derives the entity type and instance for query results automatically. The construction process solves query interpretation problems and improves query efficiency e.g. by rejecting semantically incorrect interpretations. The semantics of query results are thus also clear and accurate. This facilitates the formulation of complex queries which become simpler and more comprehensible than corresponding RA expressions. The declarativeness of the E-TRC increases its user-friendliness. The expressive power of E-TRC is of the level of Quel. Disjunctions are allowed in the predicates of E-TRC. This increases its expressive power and extends optimization beyond conjunctive queries. Because the query construction process constructs RA expressions, they can be optimized and evaluated by existing well-tested RDBMSs. The query construction process defines the semantics of the E-TRC language formally in terms of equivalent RA expressions. The entity types provide the users with a high-level and natural view of the database. The users express all query and update transactions in terms of the entity types. This simplifies the expression of complex transactions considerably w.r.t. conventional systems. The interface exploits this knowledge in an effective way.

Static and dynamic aspects of goal-oriented concurrency control

A new correctness criterion for schedules of update transactions is proposed, which captures users' intended changes to the database. This is motivated by the observation that traditional serializability may lead to anomalies by not taking into account semantics related to such intended changes. The alternate criterion —goal-correctness — is orthogonal to serializability, and is based on realizing goals associated with each transaction. The problems involved in goal-oriented concurrency control are first identified in a general framework. The analysis suggests that this approach is practical only for restricted transaction languages where goals can be inferred and manipulated efficiently. One such language is then considered, capturing a class of updates of practical interest. For this language, it is shown that goal-oriented concurrency control is tractable and compares favorably to serializability with respect to complexity: testing goal-correctness takes polynomial time, while testing serializability is NP-complete. The set of schedules which are correct with respect to the two criteria are incomparable. Thus, goal-correctness may allow increased concurrency. The results highlight the feasibility and advantages of goal-oriented concurrency control in restricted frameworks. The paper also discusses the dynamic aspects of goal-oriented concurrency control; in particular, an optimistic approach to the dynamic generation of goal-correct schedules is presented.

Efficient execution of read-only transactions in replicated multiversion databases

Multiple versions of data are used in database systems to increase concurrency. The higher concurrency results since read-only transactions can be executed without any concurrency control overhead and, therefore, read-only transactions do not interfere with the execution of update transactions. Availability of data in a distributed environment is improved by data replication. We propose a protocol for managing data in a replicated multiversion environment, where execution of read-only transactions or queries becomes completely independent of the underlying concurrency control and replica control mechanisms, and the data availability for read-only transactions increases significantly since they can be executed as long as any one copy of the object is available in the system. In order to validate the feasibility of our approach, we developed a simple prototype to measure the performance improvement in the response times of queries. The results clearly establish the viability of the approach as a useful paradigm for the design of efficient and fault-tolerant distributed database systems.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

On the design of replicated databases

Providing data availability is a prime consideration in database systems. Replication is a natural paradigm to increase the data availability in distributed databases. In this paper we focus on how to allocate the copies of a replicated database system to increase its fault tolerance. Our discussion is based on two intuitively appealing notions of availability, the group and the update availability requirement. Whereas the group represents the highest level of availability for read-only transactions, the update availability requirement is a very appealing notion of availability for update transactions. Both positive and negative results are reported. We show that there is no replica control algorithm that meets both requirements simultaneously and discuss the limitations of voting strategies in realizing groups. We demonstrate how data replication can be performed to provide significant improvements in fault tolerance. Algorithmic aspects of this issue are discussed.

Efficient and flexible methods for transient versioning of records to avoid locking by read-only transactions

We present efficient and flexible methods which permit read-only transactions that do not mind reading a possibly slightly old, but still consistent, version of the data base to execute without acquiring locks. This approach avoids the undesirable interferences between such queries and the typically shorter update transactions that cause unnecessary and costly delays. Indexed access by such queries is also supported, unlike by the earlier methods. Old versions of records are maintained only in a transient fashion. Our methods are characterized by their flexibility (number of versions maintained and the timing of version switches, supporting partial rollbacks, and different recovery and buffering methods) and their efficiency (logging, garbage collection, version selection, and incremental, record-level versioning). Distributed data base environments are also supported, including commit protocols with the read-only optimization. We also describe efficient methods for garbage collecting unneeded older versions.

Conceptual level concurrency control of relational update transactions

The concurrency control problem is examined for transactions as they appear at the conceptual level in a concrete database model. Specifically, a class of widely accepted update transactions in relational databases is studied with respect to concurrency control. It is shown how static serializability testing, as well as dynamic scheduling, can be improved by making use of the semantic information available at the conceptual level on transactions and database constraints.

A new multiversion concurrency control algorithm for partial update transactions and bulk retrieval transactions

With the advanced use of databases, on-line and batch transactions are executed concurrently. Then a demand arose for the concurrency control method which can realize the excellent on-line transaction processing independently of the batch transactions. The multiversion concurrency control is suitable for such a requirement. This paper proposes a new multiversion concurrency control algorithm which has a high concurrency between the partial update transactions, which is a typical on-line transaction, and the bulk retrieval transactions, which is a typical batch transaction. The algorithm is based on a mixture of an improved two-version, two-phase locking and the multiversion time stamp ordering. It is shown as simulation results that the proposed algorithm is better than the traditional ones in terms of the completion ratio of the partial update and the completion ratio of the bulk retrieval, independently of the number of partial updates executed concurrently. It is shown also that the algorithm is better than the traditional ones in such an environment since the data contention between partial update transactions is high.

Simplification rules and complete axiomatization for relational update transactions

Relational update transactions consisting of line programs of inserts, deletes, and modifications are studied with respect to equivalence and simplification. A sound and complete set of axioms for proving transaction equivalence is exhibited. The axioms yield a set of simplification rules that can be used to optimize efficiently a large class of transactions of practical interest. The simplification rules are particularly well suited to a dynamic environment where transactions are presented in an on-line fashion, and where the time available for optimization may consist of arbitrarily short and sparse intervals.

A scheme for increasing availability in partitioned replicated databases

We present a dynamic voting scheme that enhances the availability of data in a replicated database system in the presence of network partitions. When a network partition occurs, the distributed system splits into several groups such that nodes in one group cannot communicate with nodes in another group. In such situations, the proposed scheme maintains consistency of replicated data by imposing the following constraint: a node can update only if its current enclosing group is a majority of its previous enclosing group. Thus, the number of nodes that constitutes a majority is reduced dynamically upon successive partitions. A distinguishing aspect of the proposed scheme is that each node maintains its view of the network and keeps track of partitions that occur between successive update transactions. This technique minimizes the situations in which no group can perform an update. Another interesting feature of the proposed scheme is that it does not require failures and repairs of links and nodes to be detected instantaneously by all nodes. Different nodes may have different views of the network at any time. This introduces complications when partitions merge, and we present a merge protocol that takes this into account. We also outline a correctness proof for our proposed scheme.

Performance analysis of dynamic locking with the no-waiting policy

A transaction processing system with two-phase dynamic locking with the no waiting policy (DLNW) for concurrency control is considered. In this method, transactions making conflicting lock requests are aborted and restarted rather than blocked, thereby eliminating blocking delays (and deadlocks), but making it susceptible to cyclic restarts. Cyclic restarts are dealt with by delaying the restart of a transaction encountering a lock conflict or replacing it with a new transaction. Analytic solution methods for evaluating the performance of the variants of the DLNW method are described. The analytic methods, validated against simulation and shown to be acceptably accurate, are used to study the effect of the following parameters on system performance: transaction size and its distribution, degree of concurrency, the throughput characteristic of the computer system, and the mixture of read-only query and update transactions. A comparison of the DLNW and dynamic locking with waiting (DLW) methods shows that DLW provides higher throughput than DLNW, except when there is no hardware resource contention and conflicted transactions can be replaced by new transactions. The DLNW method outperforms the time-stamp ordering method, as observed from simulation results as well as case by case analyses of possible scenarios. >

Optimal Reorganization Policies for Stationary and Evolutionary Databases

The problem of determining optimal reorganization policies for databases which employ file structures with overflow chaining is studied. The dynamics of the file performance driven by update transactions and reorganizations is formulated as a stochastic control model which incorporates micro-level design parameters of the physical file structure. Various simplifying assumptions employed in past research are relaxed in the model. Polynomial time procedures for solving the optimization models are developed for two cases: when the file size is stationary as in the steady-state and when the file size evolves stochastically with a nonlinear trajectory. The model and the solution procedures are applied to an ISAM file revealing the effectiveness of the solution procedures and the relationship between file design parameters and the optimal policy.

Data replicas in distributed information services

In an information distribution network in which records are repeatedly read, it is cost-effective to keep read-only copies in work locations. This paper presents a method of updating replicas that need not be immediately synchronized with the source data or with each other. The method allows an arbitrary mapping from source records to replica records. It is fail-safe, maximizes workstation autonomy, and is well suited to a network with slow, unreliable, and/or expensive communications links. The algorithm is a manipulation of queries, which are represented as short encodings. When a response is generated, we record which portion of the source database was used. Later, when the source data are updated, this information is used to identify obsolete replicas. For each workstation, the identity of obsolete replicas is saved until a workstation process asks for this information. This workstation process deletes each obsolete replica, and replaces it by an up-to-date version either promptly or the next time the application asks for this particular item. Throughout, queries are grouped so that the impact of each source update transaction takes effect atomically at each workstation. Optimizations of the basic algorithm are outlined. These overlap change dissemination with user service, allow the mechanism to be hidden within the data delivery subsystem, and permit very large networks.

A study of the behavior of the read: Write ratio under two-phase locking schemes

A number of synchronization techniques have been proposed for managing the concurrent execution of transactions. The performance study of these mechanisms has been done mainly with update transactions but the behavior of read-only transactions and the effect of read: write ratio have not been looked at in depth. Since the effect of read-only transactions and read: write ratios seems significant we have employed detailed and realistic simulation models of two two-phase locking schemes to study the behavior of read-only transactions and the effect of read: write ratios on several aspects of database systems. We report that the effect of read: write ratio is not as global as usually expected. We show that it is not the write-only transactions but the read-only transactions that require special treatment for achieving improved throughput.

Properties and update semantics of consistent views

The problem of translating view updates to database updates is considered. Both databases and views are modeled as data abstractions. A data abstraction consists of a set of states and of a set of primitive update operators representing state transition functions. It is shown how complex update programs can be built from primitive update operators and how view update programs are translated into database update programs. Special attention is paid to a class of views that we call “consistent.” Loosely speaking, a consistent view is a view with the following property: If the effect of a view update program on a view state is determined, then the effect of the corresponding database update is unambiguously determined. Thus, in order to know how to translate a given view update into a database update, it is sufficient to be aware of a functional specification of such a program. We show that consistent views have a number of interesting properties with respect to the concurrency of (high-level) update transactions. Moreover we show that the class of consistent views includes as a subset the class of views that translate updates under maintenance of a constant complement. However, we show that there exist consistent views that do not translate under constant complement. The results of Bancilhon and Spyratos [6] are generalized in order to capture the update semantics of the entire class of consistent views. In particular we show that the class of consistent views is obtained if we relax the requirement of a constant complement by allowing the complement to decrease according to a suitable partial order.

Transactions for concurrent object-oriented programming systems

Transactions for concurrent object-oriented programming systems

The Experimental Study of CODASYL Database Administration at the Secondary Storage Level

Five issues of CODASYL database administration are investigated in this paper. First, the effects of CODASYL storage record placement strategies on the cost of query and update transactions are investigated. Second, the cost of CODASYL database usage pattern is analysed and it is shown that for different usage patterns, different optimal storage structures result. Third, the effects of some changes on an optimal storage structure which the designer may consider desirable are investigated. Fourth, the benefits of clustering storage records which are hierarchically related by CODASYL sets are analysed and shown to decrease with the increase in the levels of the hierarchy. Finally, the effects of statistical variation that may be expected in data populations on the cost of query and update transactions are examined.

Equivalence and optimization of relational transactions

A large class of relational database update transactions is investigated with respect to equivalence and optimization. The transactions are straight-line programs with inserts, deletes, and modifications using simple selection conditions. Several basic results are obtained. It is shown that transaction equivalence can be decided in polynomial time. A number of optimality criteria for transactions are then proposed, as well as two normal forms. Polynomial-time algorithms for transaction optimization and normalization are exhibited. Also, an intuitively appealing system of axioms for proving transaction equivalence is introduced. Finally, a simple, natural subclass of transactions, called strongly acyclic, is shown to have particularly desirable properties.