Oracle on NT or Unix for 24x7

[elintrend]

Hi all,

We are in the transition phase that should we go Unix or NT for providing 24x7 for our company. Currently, we are running Oracle8.0.6 on NT4.0-sp6a and once a while, NT needs to reboot. Is there any statistics number (report) can be some kind of evdience for my big boss that we should go Unix to have 0 down time as he wants. Where can I find them? What are pros and cons for NT and Unix...
I have done some research from metalind "high availability....' but those results seem not strong enough to convince my BIG Boss yet

Please advise.

Thanks

[jgmagnus]

maybe others will disagree, but I have found Unix VERY stable and reliable. That is the platform I would recommend.
I have no links for documentation on this however.
Shouldn't be too hard to find on the web though.

You almost never need to reboot the machines.

- Magnus

[marist89]

I'm a Solaris bigot, so I say Unix.

The differences are documented many places on the internet. Go to google.com and type in unix vs nt and you will get every possible combination of why you should go with one vs. the other.

Personally, I like to rely on TPC benchmarks for speed comparisons. http://www.tpc.org

As far as availability, you could go either way. NT setup by a good NT Admin is more reliable than any Unix flavor setup by a good NT admin.

[sambavan]

I would vote for Unix.

Sam

[elintrend]

Unix is my bet too.

Thanks, I will go google.com to find something.

[jrpm]

There is not a platform that does not need an occasional reboot, but not necessarily for the OS.

There is a glitch in Oracle that requires a reboot periodically. Depending on platform, release and version. For example, I believe Oracle 8 on Solaris needs to be reset every 27 days. The problem is the elapsed time since startup counter within Oracle. It overflows. This is true for most Oracle releases, but the time necessary varies.

It has been my experience that your best bet for 24x7 is a clustered or parallel server environment, whether Unix or Windows/NT.

And keep your fingers crossed.

[pando]

not 27, 249 days if I rememeber right and it was a bug not because of OS, I have seen a HP-UX server up for over 3 years with just 2 reboots because of kernel patching

24x7 not necessary involves Cluster, if we have cluster then we would have 0 downtime no matter the OS. OPS would be one of last choices because not every application works for OPS

NT kind sux, I rememeber we had a customer who had a 24x7 enviroment in NT, no matter what he did he had to reboot every 2.5 ~ 3 months time, it always had problem and of course noone would know what is the problem because the error messages in NT are so descriptive

[sambavan]

Originally posted by jrpm
There is not a platform that does not need an occasional reboot, but not necessarily for the OS.

There is a glitch in Oracle that requires a reboot periodically. Depending on platform, release and version. For example, I believe Oracle 8 on Solaris needs to be reset every 27 days. The problem is the elapsed time since startup counter within Oracle. It overflows. This is true for most Oracle releases, but the time necessary varies.

This 249 day issue was already solved and there was a patch for it. I have couple of Solarises that had been running for the last 2yrs, without any glitch, other than that of taking it down to apply some kernel patches last week! I had one Dec running for 5yrs at a strech!!!

You will always encounter patches for the oracle and application of those patches are at times critical and those times cannot be compromised. On the windows, you sure can expect the Blue screen to pop up as it wishes Bill needs to work on this

Sam