I was wondering if any of you notice true relationship between server hardware and uptime? Can hardware quality come into direct relationship with uptime? Can you compare such a relationship at all?

Best,

I believe hardware quality plays a role in the uptime over the longer term lifetime of the server. I believe though that software and software configuration plays a bigger role in uptime.

Mathematically speaking, yes, uptime will (partially) depend on the quality of the hardware. All hardware is (or should be) characterised by MTBF (mean time between failures) and MTTR (mean time to repair) figures.

Absolutely, hardware quality plays a role. But as others have already mentioned it is just part of the equation. Properly installing, configuring, and maintaining the OS is a big part of reliability (uptime) but not the be all end all. How many large companies take a no frills, white box, desktop system, throw a server OS on it and call it a server? Yup, that just isn't done. When your data is important and your business depends on accessing it you don't bet it on that type of equipment.

We've only ever used Intel based web servers - but at home I can tell you that my Intel and AMD systems have run equal uptimes (on average) over the past couple months. Intel desktop gaming machine and AMD domain controller. I'm upgrading that to an Intel domain controller - but only for cooling reasons as it is in a closet.

As for Hard Drives I only use Western Digital or Samsung. Always have used 3Com network cards and Lite-On optical drives. Kingston/Crucial/OCz have been RAM choices and integrated video or nVidia for display adapters.

In my entire experience I've personally only ever had problems with AMD processors - but only because they are so fragile and volitile that I wouldn't use them in a web server, hence the Intel Only remark in my first sentence. I do use AMD processors in basic desktop machines though - something that won't be on 24/7 and won't be used for much more than web surfing/e-mail/word processing.

Hope that helps.

I would have to say that hardware does play a role in uptime. "Bad" hardware could take too much time to repair and may not be able to handle the software that "good" hardware can. However, I would hope that a host would have both "good" hardware and have the knowledge to monitor their systems to keep downtime at a minimum.

I just deployed a bunch of HP DL385's for a client which use AMD Opteron processors and they are very nice servers. Only time will tell if they hold up as well as the previous generation which was Intel based.

Hardware does play a factor somewhat in terms of reliability. It can also prevent crashing if the person isnt managing their server well and they got a dual cpu so they're able to stop a infinite loop killing the server. I'd say though generally hardware isn't causing downtime but rather the admin running the server.

I should clarify that we've only ever used Intel based servers - but I'm not sayin' anything bad about AMD based servers. In my personal experience, however, I have to say for a server environment I prefer Intel over AMD. However that experience is "outdated" as the latest AMD processors I own are AthlonXP's.

I have a "consumer class" computer (Dell Optiplex) working as a print server/file archive at the office, I consider it a cheap PC - Yesterday the CPU fan burned out and I wasn't here, today I spent the first 4 hours of the day reinstalling the OS and restoring from an old Ghost image.

If it had been an IBM xSeries or an HP Proliant, it would have had redundant cooling; hence no CPU/system failure.

Yes hardware makes or breaks your business. Power supplies that aren't redundant, will fail. Same goes for cooling, RAM, drives, etc. Ever since the big server manufacturers introduced Chipkill memory I can't imagine going back --

The manufacturers have pretty much assured you 100% uptime as long as you can keep electricity flowing into the machine, and know what you're doing on the O/S & Security side of things.

I still have a Proliant 3000 with 2 x P2/400's which I've had on 24/7 for the last 5+ years. Original drives in it, haven't done anything to the box. I can't remember the last white box I had that lasted more than 2 years. Now that we're talking about hardware, how about the Compaq Evo laptop I've been running around with for 3 years, 100% original as well..

You know brand name professional equipment is now within everyone's reach, last time I checked HP was financing servers, not a bad idea if you want to be 100% sure your equipment will be running. Consider that if you have a Proliant DL580 for example, and you've got 4 X 18.2GB drives in it, it's Saturday 11 PM, and your drive fails -- you will immediately get a SMS to your cell phone from the Insight board (if you set it up right), you can then call HP and they WILL go to the Datacenter and replace that drive within 4 hours (if you have the 4 hr contract, it's worth the €€).

Yes, buy brand name, even if it's "slower" on paper. (That doesn't mean getting SC series Dells either, puke..)

I still say that your operating system is far more important than the hardware itself. I have a nix box which although I dont run huge processes on it anymore, was up for nearly 18months before I replaced the power supply. I had the same system with a winblows setup and the uptime was horrible. Blue screens and constant system faliure. Although this might have been in part my bad setup skills, it has put me off windoze for life and I know only operate off the operating system which come from unix-land

~Equinox

Totally agreed. That Proliant we're referring to has been running Red Hat since '00. I've got 7.3 on it now and it's going to stay that way.

I love *nix!

(but w2K3 is great too if done right)

Which in turn makes the server admin very important.

Which in turn makes my salary quite healthy...hmmm...who said all circles had to be vicious?

Having recently suffered from a PSU failure then yes hardware can affect uptime.

I think in the case of a server its worth making sure you have the hardware for the job.