No abstract

Though in general, current database systems ade quately support application development and opera tion for online transaction processing (OLTP), in creasing complexity of applications and throughput requirements reveal a number of weaknesses with respect to the data model and implementation techni ques used. By presenting the experiences gained from a case study of a large, high volume stock trading sys tem, representative for a broad class of OLTP ap plications, it is shown, that this particularly holds for dealing with high frequency access to a small number of data elements (hot spots). As a result, we propose extended data types and several novel mechanisms, which are easy to use and highly increase the expressional power of transaction oriented programming, that effectively cope with hot spots. Moreover, their usefulness and their ability to increased parallelism is exemplified by the stock trading application. Functional and operational characteristicsDatabase systems, especially those of the relational type, are regarded as the basic utility to manage the operational data of an enterprise [1]. Data inde pendence and the system-controlled data integrity facilitate the integration and extension of increasingly complex applications, especially in the online transac tion processing (OLTP) field. Though in general, cur rent database systems provide suitable means for OLTP application development and operation, in creasing complexity of applications and throughput requirements reveal a number of serious weaknesses in both, the data model and the implementation tech niques used [2]. They mainly arise from the inap propriate support for dealing with high frequency ac cess to a small number of data elements present in the majority of applications. Though several proposals, like the introduction of field calls into IMS/FP [3], try to eliminate that problem, we claim that those primitives still do not provide an adequate solution for a wide range of real-life applications, and will describe several novel mechanisms to cope with hot spot data elements. Our claims stem from observations made in a case study [4], which included a prototype implementation of a large high volume stock trading system, using the features of a standard relational database system. To make our point, the remainder of this section will out line the functional and operational characteristics of this application. Section 2 will provide some insight into the actual implementation, the problems caused by the use of standard data management primitives and the tricks required to circumvent them. Section 3 then will draw some general conclusions from these observations and come up with the definition of novel data management requirements. Thereafter, in section 4, we propose practical solutions for the management of hot spot data, and finally, in section 5, we put these functional extensions into the perspective of an en hanced DBMS. Fig. 1 illustrates the general structure of the com puterized stock trading system, which basically provides three type...

Though in general, current database systems adequately support application development and operation for online transaction processing (OLTP), increasing complexity of applications and throughput requirements reveal a number of weaknesses with respect to the data model and implementation techniques used. By presenting the experiences gained from a case study of a large, high volume stock trading system, representative for a broad class of OLTP applications, it is shown, that this particularly holds for dealing with high frequency access to a small number of data elements (hot spots). As a result, we propose extended data types and several novel mechanisms, which are easy to use and highly increase the expressional power of transaction oriented programming, that effectively cope with hot spots. Moreover, their usefulness and their ability to increased parallelism is exemplified by the stock trading application.