Requires Free Membership to View
Benchmarks are very difficult to come by and to trust. One can look at
the TPC benchmarks, but you'll have to take these
figures with a grain of salt. What is a benchmark today will be old
news tomorrow. If you pay attention to these benchmarks, you will find
Oracle on top one day, SQL on top the next, and so on. And any
benchmarks typically are far removed from "real world" situations. The
minute you throw Oracle on a Windows server and are happy with its
performance, I can come up with a Linux server, with more memory,
faster CPUs, and faster disk that will outperform your original
configuration. And vice-versa. Personally, if you are really worried
about performance, then you might want to look at running Oracle on a
high-end Unix platform.
But if you want generic guidelines, then my rules of thumb are as follows. First, Oracle on Linux is typically faster than Oracle on Windows, provided that your Linux server is not running its GUI interface. The GUI interfaces for the OS can slow things down quite a bit. Second, in either case, make sure that you are using SCSI disks (or even Serial ATA disks). Just stay away from IDE disks if at all possible. IDE disks can't satisfy concurrent disk requests. It has to single stream disk requests and this can slow things down.
And it has been my experience that Linux is more stable and more secure
than Windows. Windows 2000 Server and Advanced Server have become much
more stable so the differences might not be that much. I'd rather
deploy Oracle on Linux, but I've often deployed Oracle on Windows as
well with great success.
This was first published in October 2003
Join the conversationComment
Share
Comments
Results
Contribute to the conversation