Databases Reference
In-Depth Information
Of the operating systems in our list LINUX is the only open-source operating sys-
tem. LINUX is a relative newcomer on the OS scene, but the rapid adoption of LINUX
by the open-source community, as well as the large range of developers for LINUX from
the open-source community, has allowed LINUX to mature rapidly. As of 2007,
LINUX remains an emerging platform for enterprise-scale database processing, though
it has seen rapid adoption for departmental level applications, and is gaining strength
rapidly. Major products, such as MySQL, DB2, and Oracle, all support LINUX. AIX,
Solaris, and HP are enterprise-scale operating systems. All of these have lost some mar-
ket share to LINUX in recent years though HP appears to be the hardest hit.
All of the major database vendors support execution on Windows. However, SQL
Server produced by Microsoft does not support any of the UNIX platforms. Therefore,
if SQL Server is your database of choice, your choice of operating system is made for
you—Windows is your only option.
Another factor in operating system choice is the “one-stop shop” factor.
3
Purchasing
multiple components from a single vendor reduces the complexity of managing and ser-
vicing the complete database solution. For example, if you purchase the operating system
and the database from the same company, any questions or problems relating to file cach-
ing have a clear target for who should be able to answer your questions. When the data-
base and operating system are purchased from distinct vendors there are more vendors to
work with and vendors can more easily evade responsibility for problems by pointing a
finger at another component in the system. Purchasing an operating system and a data-
base from the same vendor is currently possible with SQL Server on Windows, DB2 on
AIX, and Teradata on NCR (strictly a warehousing solution).
13.6
Storage Systems
Throughout this topic we have discussed the value of physical database design in reduc-
ing input/output (I/O) requirements for database systems. Intelligent and careful use of
indexes, materialized views, data clustering, and range partitioning can all help reduce
the I/O burden. However, despite the very best application of physical database design
features, unless all the database data can be cached simultaneously in main memory, I/O
will be required. Most databases used for online transaction processing (OLTP) and
decision support systems (DSS) the I/O requirements remain very significant. It is com-
mon for databases larger than 500 GB pre-load data to be I/O bound
4
when they go
into operation. This means that although the user has paid for significant CPU capacity
they aren't able to make use of it, because the CPU is waiting for I/O. Database vendors
3
Or occasionally known as the “one throat to strangle” factor. Buying multiple components from
a single vendor reduces the number of vendors to deal with when there are questions or defects.
4
Author's experience.
