For many years, TIBCO (the Information Bus Company) has pioneered the use of Publish/Subscribe—a form of push technology — to build flexible, real-time loosely-coupled distributed applications. Today, Publish/Subscribe is used by 300 of the world's largest financial institutions, deployed in 6 of the top 10 semiconductor manufacturer' factory floors, utilized in the implementation large-scale Internet services like Yahoo, Intuit, and ETrade, and chosen by many of the world's leading corporations as the enterprise infrastructure for integrating disparate applications. In this paper, we will:

Adaplex is an integrated language for programming database applications. It results from the embedding of the database sublanguage Daplex in the general-purpose programming language Ada [1]. This paper describes the design of DDM, a general-purpose distributed database management system implemented in Ada that supports the use of Adaplex as interface language. There are two novel aspects in the design of this system. First, this is the first full-scale distributed database system to support a semantically rich, functional data model. DDM goes beyond systems like Distributed INGRES and R*(which are based on the relational technology) in providing advanced data modeling capabilities and ease of use. Second, this is the first full-function distributed DBMS designed to be compatible with the Ada programming environment. The coupling between Ada and Daplex has been achieved at the expression level which is much tighter than the statement level integration attained in previous systems. This tight coupling poses new implementation problems but also creates new opportunities for optimization. The current paper highlights the Adaplex language and discusses innovative aspects in DDM's design that are intended to meet the dual objectives of good performance and high data availability.

In a large integrated database, there often exists an “information hierarchy,” where both raw data and derived data are stored and used together. Therefore, among update transactions, there will often be some that perform only read accesses from a certain (i.e., the “raw” data) portion of the database and write into another (i.e., the “derived” data) portion. A conventional concurrency control algorithm would have treated such transactions as regular update transactions and subjected them to the usual protocols for synchronizing update transactions. In this paper such transactions are examined more closely. The purpose is to devise concurrency control methods that allow the computation of derived information to proceed without interfering with the updating of raw data. The first part of the paper presents a proof method for correctness of concurrency control algorithms in a hierarchically decomposed database. The proof method provides a framework for understanding the intricacies in dealing with hierarchically decomposed databases. The second part of the paper is an application of the proof method to show the correctness of a two-phase-locking- based algorithm, called partitioned two-phase locking, for hierarchically decomposed databases. This algorithm is a natural extension to the Version Pool method proposed previously in the literature.

This paper presents an efficient scheme for eliminating conflicts between distributed read-only transactions and distributed update transactions, thereby reducing synchronization delays. The scheme makes use of a multiversion mechanism in order to guarantee that distributed read-only transactions see semantically consistent snap-shots of the database, that they never have to be rolled-back due to their late arrival at retrieval sites, and that they inflict minimal synchronization delays on concurrent update transactions. Proof that the presented scheme guarantees semantic consistency is provided. Two important by-products of this scheme are that the recovery from transaction and system failures is greatly simplified and the taking of database dumps also can be accommodated while leaving the database on-line.

We describe a reliability algorithm being considered for DDM, a distributed database system under development at Computer Corporation of America. The algorithm is designed to tolerate clean site failures in which sites simply stop running. The algorithm allows the system to reconfigure itself to run correctly as sites fail and recover. The algorithm solves the subproblems of atomic commit and replicated data handling in an integrated manner.

Adaplex is an integrated language for programming database applications. It results from the embedding of the database sublanguage DAPLEX in the general purpose programming language Ada. This paper provides an overview of the DDM: a distributed database manager (DDM) that supports the use of Adaplex as an interface language. The important technical innovations we have incorporated in the design of this system include:1. An advanced data model that captures more application semantics than conventional data models.2. Support for flexible data distribution options that improve locality of reference and efficiency of query processing.3. Extensive query optimization that combines compile time access path optimization with run time site selection.4. Efficient transaction management that reduces transaction conflicts and improves the resiliency of replicated data.5. Robust, incremental recovery management that provides for automatic recovery from certain "catastrophic" failure conditions.

This paper describes the implementation level design of an integrated concurrency control and recovery scheme based on the maintenance of multiple versions of data objects in a database. The concurrency control mechanism enhances parallelism by eliminating interference between retrieval and update transactions. The recovery mechanism permits efficient transaction and system recovery by keeping before-images of data objects at the page (block) level. This paper addresses the key technical problems in the implementation of such an integrated scheme. We present an efficient garbage collection algorithm for reclaiming storage space used by old versions of data objects that will no longer be accessed. We also propose an on-line backup algorithm that will permit the backup procedure to run in parallel with regular transactions. This integrated concurrency control and recovery scheme is being implemented in the LDM: the local database manager component of a distributed database management system, now being developed by Computer Corporation of America, that will support the ADAPLEX database application programming language [Chan81a, Smith81].

The estimation of the cost of processing a query using a particular access path under a given physical organization has important applications in integrated database environments. When records in a file are stored in fixed-length physical blocks in secondary storage, and mechanisms are available whereby a query can be resolved without the accessing of all of the records, an important measure of the cost of using a particular access path is the number of blocks that have to be accessed in referencing the records of interest. In this paper, a general formula is derived for the expected number of blocks on which a random sample of r records from a file containing n records (which may be of arbitrary lengths, and which may extend across block boundaries) will reside. The specialization of this formula to the case of fixed-length records is discussed. An approximation to this formula which is highly accurate for a wide range of parameters and which can be computed very efficiently is also provided.

We address the problem of automatically adjusting the physical organization of a data base to optimize its performance as its access requirements change. We describe the principles of the automatic index selection facility of a prototype self-adaptive data base management system that is currently under development. The importance of accurate usage model acquisition and data characteristics estimation is stressed. The statistics gathering mechanisms that are being incorporated into our prototype system are discussed. Exponential smoothing techniques are used for averaging statistics observed over different periods of time in order to predict future characteristics. An heuristic algorithm for selecting indices to match projected access requirements is presented. The cost model on which the decision procedure is based is flexible enough to incorporate the overhead costs of index creation, index storage and application program recompilation.