On Mon, 09 Nov 2009 15:11:32 +0200, Stefan Viljoen wrote:

> Jim Henderson wrote:
> 
>> On Mon, 09 Nov 2009 09:42:27 +0000, Invisible wrote:
> 
>> See my anecdotal evidence in reply to Stefan.  Two identical units, one
>> powered on and off regularly, one that was left on 24x7.  Guess which
>> one failed?  Not the one that has been turned on and off regularly for
>> 5 years now.  The one that was plugged in and running for 3 years
>> solid.
> 
> Hmm... well, no contest there. I guess there can be other factors
> involved besides on / off frequency then? My personal experience with
> server-grade hardware I've worked with differs though. It does seem (at
> least in my personal experience) that turning HDD's on and off less
> makes them last longer. But as you say, anecdotal. :)

Yeah, that's the problem with anecdotal evidence, of course.  Server 
grade hardware that I've worked with has varied - I had a Compaq 
Prosignia 4/66 with a RAID controller in it, and it had heat problems - I 
occasionally would lose two drives in the array.  The system was used in 
my home, so I didn't have the kind of cooling you'd have in a computer 
room.

>>> PS. I've yet to find a consumer RAID controller which actually works
>>> properly. They all seem to be hopelessly unreliable. And most of them
>>> are software RAID anyway; the controller is a normal IDE or SATA
>>> controller, and the Windows driver does all the actual RAID functions
>>> in software. It seems that only the £££ server-grade controllers
>>> actually do the job properly.
>> 
>> Software raid exists for some platforms - and for mirroring for backup
>> purposes, that's really all you need.  Heck, you could do it over iSCSI
>> with a fast enough connection (I've seen that done as well).
> 
> I've written a simple cronjob which uses bash scripting and some
> command-line PHP to automatically tar/7zip our primary server's vital
> sites and databases once every 24 hours, and then FTPs that over to our
> over server in a different datacenter automatically. Works fine too. The
> data centers are about 10km away from each other.

I operate a website for our local community council, and while the system 
is hosted by a local ISP, I take a nightly backup as well - using sshfs 
for the files and a cron job to dump the database.

> That way we can lose only one day's worth of changes, which (cause we're
> small!) usually isn't much. But I can see the point of some kind of
> "live" solution if you do a TB a day or whatever.

Yep.

Jim

Post a reply to this message