| Previous | Table of Contents | Next |
by Derek C. Ashmore
Online Transaction Processing (OLTP) applications are applications that record information about a specific business unit function. For example, most banks have OLTP applications in which bank officers can enter information about new customers or accounts for those customers. OLTP applications, by definition, obtain most input manually from an end user.
This chapter provides recommendations on how to effectively design, code, and administrate OLTP applications that use Oracle as an underlying database. Typically, OLTP applications have extensive availability requirements. For example, one of my clients administrates credit card accounts. If this system isn’t available, customers cannot use their credit cards for purchases. This can mean lost revenue and ill will from customers.
Another of my clients uses an OLTP application to record treatments for patients in a hospital. This information is used to inform nurses which patients get which treatments. Application outages cost the hospital in many ways, including vulnerability to lawsuits when prescribed treatments are missed, the ill will of patients who suffer lack of treatment, and so on. Frequently, large OLTP applications also have high availability requirements.
Typically, non-MVS platforms provide high availability capabilities via clustering technology. Clustering technologies allow UNIX, and more recently Windows NT servers, to be tightly coupled enough to share memory and disk. In other words, each node or machine in the cluster can access the same set of disks. Furthermore, machines, which are clustered, often can share memory.
Oracle supports clustering technologies with an add-on called the Oracle Parallel Server (OPS). This add-on allows you to have multiple instances (one on each node) providing access to the same database (the same set of data files). This means that each instance has access to the same set of tables, indexes, sequences, and so on. OPS is often used with OLTP applications because of the high availability requirements, so this chapter will provide an overview of clustering technologies and OPS.
Typically, OLTP applications have large numbers of users (over 1,000), mainly small targeted transactions that involve small result sets for most queries (keyed updates, deletes, and selects), and a large proportion of writes. Depending upon the application, this chapter will provide recommendations on how to effectively design and configure databases for this type of load. Furthermore, we will provide recommendations on how to effectively structure database access within applications to make them more efficient and less resource intensive.
OLTP applications typically have large numbers of users, so most are hosted on RISC platforms running some variety of UNIX. At the time of this writing, Windows NT is simply not scalable enough to handle these types of applications. This chapter will give you recommendations on how to configure UNIX servers to support OLTP applications. Discussion of performance tuning Windows NT is purposely omitted from this chapter because Windows NT cannot yet effectively support OLTP applications with thousands of users.
This section will provide an overview of OLTP database design and configuration issues, including the use of OPS to meet typical high availability requirements of OLTP applications. We will also discuss database design and configuration issues for non-OPS applications, as well as UNIX configuration considerations.
| Previous | Table of Contents | Next |