| Previous | Table of Contents | Next |
by Meghraj Thakkar
Very large databases (VLDBs) are in the range of terabytes and before long will become petabytes. The task of managing and maintaining databases increases at least proportionally, if not exponentially, with their size. With the increase in the data, the tasks such as backups, recovery, data loads, purging, and data reorganization become more and more complex and time consuming. Most of the customers who have large databases installed usually have additional requirements such as high availability, high response time, and high scalability.
There are several design issues that need to be considered when creating very large databases in order to take full advantage of the features provided by Oracle8. This will also have a major impact on the management and performance of VLDBs.
This chapter addresses the various issues related to designing and managing VLDBs under Oracle8 or migrating them from Oracle7. We will also discuss the different features provided by Oracle8 from a VLDB perspective. Some of these features are
You should be very careful when dealing with high-end applications. These applications tend to test the database limits, and one has to be very creative to make these applications run optimally. High-end applications have one or more of the following characteristics:
Most of the high-end database environments are implemented by using mainframes and non-relational databases, because even though relational database systems are easy to use and manage, they are not able to handle the types of load imposed by high-end systems.
Oracle8 provides several features that allow it to work in an acceptable manner in such high-end environments. These features are discussed in the following sections.
Oracle8’s data partitioning features can simplify the demanding tasks involved when dealing with large amounts of data. Table 33.1 shows some problems faced in VLDB systems and how Oracle8 attempts to resolve those problems.
| Problem | Oracle8’s Solution |
|---|---|
| Performing full table scans on a large table can degrade performance. | Oracle8 provides the ability to perform parallel DMLs and query individual partitions. The optimizer is “partition aware” and eliminates searches from partitions that do not apply to a particular SQL query. |
| High availability is required in a system. Mission-critical applications depend on atable that is not available due to media failure. | Oracle8 allows you to recover at the partition-level and minimizes the downtime. |
| VLDBs are difficult to manage. | Oracle8’s partition-level management simplifies a lot of administrative tasks. |
Oracle8 and Net8 provide four possible solutions to achieve more concurrent connections per Oracle8 instance:
In VLDB systems, there are usually several tasks and applications that can benefit from deferring the execution of low-priority transactions. Third-party solutions like Transaction Processing Monitor (TP Monitor) or Message-Oriented Middleware (MOM) have problems because they are not able to protect the deferred transactions against system failure, and there is additional overhead associated with them. Oracle8 provides a new feature—advanced queuing—which is a database-centric solution that protects deferred transactions.
Oracle8 provides several enhancements to parallel query:
Oracle8 provides several enhancements to Parallel Server:
| Previous | Table of Contents | Next |