The discussion of hardware architectures and performance in this chapter so far has centered on ways of increasing performance by increasing available system resources such as CPUs, memory, and I/O subsystems and on parallelism that can take advantage of these resources.
Another classic way to increase hardware performance is to tune for I/O, which includes tuning for disk layout. Because disk access has the greatest latency, the main focus of I/O tuning is typically keeping what has been retrieved from disk in memory. The actual performance of retrieving that data from disks can also be improved by spreading the data evenly across multiple disks and by making sure there are enough disk controllers to transfer the data from disk onto the I/O bus and into memory. As noted previously, Oracle Database 10g now includes automatic data rebalancing across disks, greatly simplifying the day-to-day management. However, working with your hardware vendor regarding disk controller selection and configuration is still important.
Disks have typically been directly attached to systemsmore expensive systems offer fastest disk controllers and I/O. Disks are also configured in a variety of ways for redundancy, eliminating the possibility of single points of disk failure resulting in loss of access to data. As network bandwidth improved, deployments that include Network Attached Storage (NAS) and Storage Area Networks (SAN) have appeared as cost-effective alternatives.
Disk is commonly deployed in arrays, the industry standard being RAID (Redundant Array of Inexpensive/Independent Disks). You can use RAID as a part of any of the configurations we've discussed to provide higher performance and reliability. RAID disk arrays were introduced in Chapter 6 and discussed in the context of their use in high-availability scenarios in Chapter 10. Please refer to those chapters for more information about RAID disk arrays. In addition, Oracle Database 10g's Automatic Storage Management (ASM) capabilities can deliver much of the functionality of a RAID array, such as striping and mirroring, with a collection of commodity disks. ASM is further described in Chapter 5.
Keep in mind that newer higher capacity drives may allow you to store more and more data cheaply, but may not provide the access paths (e.g. spindles) necessary for demanding multiuser query performance. Proper disk configuration is a balance of capacity, availability, and performance considerations that meet your business application needs.
Oracle9i introduced table compression in the database as a means of decreasing disk storage requirements. Duplicate values in a data block are eliminated because values that are duplicated are stored in a symbol table at the beginning of the block, and all additional occurrences are replaced with a short reference to the symbol table. Compression ratios vary, but typical reductions observed in storage required in real implementations is about half that of uncompressed data.
In addition to reducing disk storage, compressed data can also be advantageous to query speed when it fits entirely into cache (instead of requiring disk access). Loading data as compressed can take about twice as long, so many companies load new data uncompressed in a partition, then compress that partitioned data as it ages.