Oracle Database Tips by Donald Burleson

Introduction to Oracle Data Caching

Leveraging RAM resources has always been one of the central tasks of the Database Administrator (DBA).  By definition, almost all databases are I/O intensive, and minimizing the expensive physical disk I/O has always been a major priority to ensure acceptable performance.  Historically, RAM has been a scarce and expensive resource, and the DBA was challenged to find the best working set of highly used data to cache on their precious RAM media.

However, RAM is quite different than other hardware.  Unlike CPU speed, which improves every year, RAM speed is constrained by the physics of silicon technology.  Instead of speed improvements, there has been a constant decline in price.  CPU speed also continues to outpace RAM speed.  This means that RAM subsystems must be localized to keep the CPU's running at full capacity.

In the 1980's, a billion bytes of RAM cost over a million dollars; whereas, today one gigabyte of high-speed RAM storage can be acquired for less than $300.00.  As a reference point, a billion bytes of RAM is equal to 1 Gigabyte. Some other pricing data points are outlined in Table 1.1 below:
 

Historically, RAM I/O bandwidth grows one bit every 18 months, making the first decade of the 21^st Century the era of 64-bit RAM technology as shown in Table 1.2:

The bandwidth information is interesting.  These numbers typically have as much to do with matching the bus width of the computer as they do with the speed of the RamSan chip.  It is also more commonly used to describe SIMM performance as 8 bit/16/32/64 bit, so it might not be a good measure of chip performance.  The data in Table 1.3 below was taken from a Kingston Memory webpage.  It was part of their ultimate guide to memory:

It is clear that the dramatic decreases in RAM prices are going to change Oracle RAC database architectures.  With RAM a once scarce and expensive resource, the Oracle DBA had to spend a huge amount of time managing Oracle memory allocation and optimization.  This is about to change.

Today 128 gigabytes of SSD (i.e. Texas Memory Systems) can be purchased for about $150,000.  By 2015, a Gigabyte of RAM is expected to cost much less.  Of course, less expensive solid state disks will mean a dramatic change in Oracle database architecture as the old-fashioned model of disk-based data management will be abandoned in favor of a cache-based approach.

According to David Ensor, Oracle tuning expert, author, and Former Vice President of the Oracle Corporation's Performance Group, the increase in CPU power has shifted the bottleneck of many systems to disk I/O. 

See code depot for complete scripts

This is an excerpt from the book Oracle RAC & Tuning with Solid State Disk.  You can get it for more than 30% by buying it directly from the publisher and get immediate access to working code examples.