I didn't get that performance comment in the article. Performance is
based almost completely on hardware and only in minor ways on the
operating system. It sounds like some Linux zealot employees have
been feeding the IT managers a line of bull.
It sounds more like they're going to Intel servers. If they did, they
are going to hit some huge scalability and performance problems.
Excellent point. I'm not sure why everything has to turn into an OS
religious debate. Sun is primary interested in selling their systems
and not their OS. If fact, I'm not sure why anyone would even have to
pay for Solaris unless maybe you buy a Sparc clone. Their real
interest is selling their Netra boxes for web services and the Sun
Fire/Enterprise boxes for mainframe sized servers.
As far as the OS, if it runs your stuff, who cares? Get whatever is
best tuned for the hardware. It's more important to have source
available at the application level than the core OS level. As long as
the OS follows an open standard, your stuff is portable.
You may have a special case where point in time recoverability is not
necessary. Every large application I've worked with requires it. It
becomes especially important anytime a new releases is being deployed.
A bad query can muck up your entire database and not be noticed until
weeks later.
There's some significant risk differences you may want to consider.
First, you only have a single generation of backup. If someone
totally screws the data, it will get synced over totally screwed, and
your company may be out of business. The only way to avoid the risk
is a long retention period with multiple generations on an off-site
rotation.
The other is the single failure point of the telecom line. If it goes
down for a couple of days so does your backup. This is probably
acceptable risk for most applications.
Tape is still far from dead. The bigger you scale up, the lower the
cost of ownership of tapes become. For one thing, you've got to power
all those hard drives (and cool them), and I'm not aware of any hard
drive auto changers. Tapes are also more durable for shipping. You
could actually crush a tape cartridge, wind it back into another
cartridge, and most likely be able to read it. Recovering data from
broken hard drives is very expensive.
Slashdot Mirror
Read "performance" not "features" Grow up
I didn't get that performance comment in the article. Performance is based almost completely on hardware and only in minor ways on the operating system. It sounds like some Linux zealot employees have been feeding the IT managers a line of bull. It sounds more like they're going to Intel servers. If they did, they are going to hit some huge scalability and performance problems.
Excellent point. I'm not sure why everything has to turn into an OS religious debate. Sun is primary interested in selling their systems and not their OS. If fact, I'm not sure why anyone would even have to pay for Solaris unless maybe you buy a Sparc clone. Their real interest is selling their Netra boxes for web services and the Sun Fire/Enterprise boxes for mainframe sized servers. As far as the OS, if it runs your stuff, who cares? Get whatever is best tuned for the hardware. It's more important to have source available at the application level than the core OS level. As long as the OS follows an open standard, your stuff is portable.
You may have a special case where point in time recoverability is not necessary. Every large application I've worked with requires it. It becomes especially important anytime a new releases is being deployed. A bad query can muck up your entire database and not be noticed until weeks later.
There's some significant risk differences you may want to consider. First, you only have a single generation of backup. If someone totally screws the data, it will get synced over totally screwed, and your company may be out of business. The only way to avoid the risk is a long retention period with multiple generations on an off-site rotation. The other is the single failure point of the telecom line. If it goes down for a couple of days so does your backup. This is probably acceptable risk for most applications. Tape is still far from dead. The bigger you scale up, the lower the cost of ownership of tapes become. For one thing, you've got to power all those hard drives (and cool them), and I'm not aware of any hard drive auto changers. Tapes are also more durable for shipping. You could actually crush a tape cartridge, wind it back into another cartridge, and most likely be able to read it. Recovering data from broken hard drives is very expensive.