Donald J. Haderle

ARIES: a transaction recovery method supporting fine-granularity locking and partial rollbacks using write-ahead logging

DB2<supscrpt>TM</supscrpt>, IMS, and Tandem<supscrpt>TM</supscrpt> systems. ARIES is applicable not only to database management systems but also to persistent object-oriented languages, recoverable file systems and transaction-based operating systems. ARIES has been implemented, to varying degrees, in IBMs OS/2<supscrpt>TM</supscrpt> Extended Edition Database Manager, DB2, Workstation Data Save Facility/VM, Starburst and QuickSilver, and in the University of Wisconsins EXODUS and Gamma database machine.

An efficient hybrid join algorithm: a DB2 prototype

A new join method, called hybrid join, is proposed which uses the join-index filtering and the skip sequential prefetch mechanism for efficient data access. With this method, the outer table is sorted on the join column. Then, the outer is joined with the index on the join column of the inner. The inner tuple is represented by its surrogate, equivalent of its physical disk address, which is carried in the index. The partial join result is sorted on the surrogate and then the inner table is accessed sequentially to complete the join result. Local predicate filtering can also be applied before the access of the inner relation through the index AND/ORing. Efficient methods for skip sequential access and prefetching of logically discontiguous leaf pages of B/sup +/-tree indexes are also presented.<<ETX>>

Parallel merging: algorithm and implementation results

Abstract An efficient parallel algorithm for merging two sorted lists is presented. The algorithm is based on a novel partitioning algorithm that splits the two lists among the processors, in a way that ensures load balance during the merge. The partitioning algorithm can itself be efficiently parallelized, allowing the solution to scale with increased numbers of processors. A shared memory multiprocessor is assumed. The time complexity for partitioning and merging is O(N/p + log N), where p is the number of processors and N is the total number of elements in the two lists. Implementation results on a twenty node Sequent Symmetry multiprocessor are also presented.

Database Role in Information Systems: The Evolution of Database Technology and its Impact on Enterprise Information Systems

During the 1960s and 1970s commercial database management systems focused on transaction and batch processing environments with an emphasis on minimizing computing resources. Relational database management systems appeared on the commercial scene in the 1980s and extended the range of applications amenable to database processing. These applications now include interactive environments as well as transaction and batch. Relational DBMSs emphasize function, ease of data access by users, and improved economics of application development at the expense of computing resources. This paper looks at how enterprise information systems have been affected by the evolution of database technology in the 1980s, emphasizing the role of relational database and forecasting expectations for the 1990s.

Parallelism with IBM's relational Database2 (DB2)

The current use of parallelism in the IBM Database2 (DB2) is discussed, along with future trends. Parallelism is necessary for relational database managers to achieve modern performance goals of high throughput for both homogeneous and heterogeneous workloads and fast response times for increasingly complex queries. A transaction appears to DB2 as a set of serially requested database operations using the Structured Query Languages (SQL). High throughput levels are achieved by multiprocessing. Multiprocessing levels in excess of a hundred are quite common. DB2 provides peer-to-peer communication, allowing applications on one DB2 computer system to access data on another DB2 computer system. That implementation is more favorable for ad hoc queries than for transactions. Parallelism plays a key role in DB2s performance, where it is most visible in intertransaction multiprocessing to achieve high throughput and, to a lesser extent, in intraquery parallelism to reduce response time.<<ETX>>