 |
|
Oracle Tips by Burleson |
Bandwidth and Oracle I/O
It should not be a surprise that the vast majority
of Oracle databases are I/O centric, serving-up billions of bytes of
information to the application layer. Even with the rapidly-falling
costs of RAM, large Oracle databases may perform millions of disk
reads per minute, and the I/O subsystem is the source of system-wide
bottlenecks.
Over the past decade, Oracle has developed
sophisticated caching mechanisms to minimize disk I/O, but managing
I/O throughput remains a major issue. There are several non-Oracle
components that are used to optimize data-centric applications:
– The RedHat Linux4.0 operating
system is specifically optimized for use by Oracle database, with
special attention to the I/O drivers.
– Several vendors now offer servers that are architected to maximize
I/O throughput. For example, the UNISS ES-7000 series of servers use
Non Uniform Memory Access(NUMA)
to achieve optimal I/O in a data-centric environment.
– SSD,
one of the greatest tools for achieving high data concurrency, is RAM
disk that allows high bandwidth and data transfer rates that are an
order of magnitude faster than the old-fashioned spinning platter
devices.
Preventing disk I/O bottlenecks has always been
one of the central tasks of the Oracle Database Administrator (DBA).
By definition, almost all Oracle databases are I/O intensive and
minimizing the expensive physical disk I/O has always been a major
priority to ensure acceptable performance. The only exceptions are
scientific applications such as Laboratory Information Management
Systems (LIMS), which read a small data set and perform CPU-intensive
calculations.
The above book excerpt is from:
Oracle RAC & Tuning with
Solid State Disk
Expert Secrets for High
Performance Clustered Grid Computing
ISBN
0-9761573-5-7
Donald K. Burleson & Mike Ault
http://www.rampant-books.com/book_2005_2_rac_ssd_tuning.htm
|
