asterisk users - Jan 2004 - Hardware for Asterisk

At 1/16/04 7:25 AM, Andrew Kohlsmith <akohlsmith-asterisk@benshaw.com> 
wrote:
>That's pure bullshit -- I use software RAID *specifically* because I
value
>my data.  I don't want to buy two hardaware RAID controllers to have one
>sit on the shelf just in case the first dies... and if the second dies 
>you're SOL because they've lasted long enough that they're no
longer
>available.  Linux software RAID is available on any Linux system and if the 
>system blows up I can put the drives in another system and *not* worry 
>about it not being detected.
Yeah, I couldn't agree more.

We originally thought hardware RAID was the way to go, and we bought a 
couple of fully loaded Dell PowerEdge 2550s with SCSI hardware RAID 5 
arrays at about $4500 a pop. We also bought a PowerEdge 600SC for around 
$900 with lots of disk space to use as a network backup machine (backing 
up the 2550s) with Linux software RAID 5. I've also had a crappy old 
desktop machine running Linux software RAID 1 for a couple of years.

It turns out that the software RAID is just as reliable (more so, in fact 
-- we have had a number of lockups on the 2550s that appear to be due to 
the hardware RAID subsystem locking up, and the software RAID machines 
have never done that, even though the backup server does more disk I/O 
than the others). The software RAID on the 600SC is faster than the 
hardware RAID in bonnie tests.

In addition, the Dell PowerEdge mailing lists are full of people with 
horror stories about their hardware RAID systems -- if that dies on mine, 
I'm screwed until I can convince Dell to come out and fix it (which they 
often won't do until they've spent hours on the phone with you trying 
various things).

We should have simply bought 4 600SCs (instead of 2 2550s and a 600SC), 
using one as a hot standby, and saved ourselves around $6000. In fact, 
we're planning on moving to that and selling the 2550s on eBay to improve 
our overall reliability. If the power supply, motherboard or RAM of a 
600SC dies, we can easily move the disks to the spare machine and be back 
up within a few minutes without relying on anyone else. In the worst case 
(RAID corruption/machine catches on fire), I'm still going to be okay, 
because I can restore from backups in a couple of hours.

The key thing to me is that at no point do we have to rely on any other 
company to get things up and running again, which is far more important 
than any putative risk of data corruption from software RAID (which I 
have not seen even under very heavy disk loads, and which I think is 
pretty much a myth these days; look at the Dell PowerEdge mailing lists 
if you think hardware RAID is more reliable -- those stories of hardware 
RAID problems from real users have scared me to the point that I'll never 
consider buying any sort of proprietary disk subsystem again).

-- 
Robert L Mathews, Tiger Technologies      http://www.tigertech.net/

"I am not able rightly to apprehend the kind of confusion of
 ideas that could provoke such a question."  -- Charles Babbage

Robert L Mathews wrote:
> At 1/16/04 7:25 AM, Andrew Kohlsmith
<akohlsmith-asterisk@benshaw.com>
> wrote:
> 
> 
>>That's pure bullshit -- I use software RAID *specifically* because I
value
>>my data.  I don't want to buy two hardaware RAID controllers to have
one
>>sit on the shelf just in case the first dies... and if the second dies 
>>you're SOL because they've lasted long enough that they're
no longer
>>available.  Linux software RAID is available on any Linux system and if
the
>>system blows up I can put the drives in another system and *not* worry 
>>about it not being detected.
> 
> 
> Yeah, I couldn't agree more.
> 
> We originally thought hardware RAID was the way to go, and we bought a 
> couple of fully loaded Dell PowerEdge 2550s with SCSI hardware RAID 5 
> arrays at about $4500 a pop. We also bought a PowerEdge 600SC for around 
> $900 with lots of disk space to use as a network backup machine (backing 
> up the 2550s) with Linux software RAID 5. I've also had a crappy old 
> desktop machine running Linux software RAID 1 for a couple of years.
[snip a couple HW/SW RAID horror stories]

Personally, I don't like any kind of "internal" RAID solution (ie.
software or a PCI card).  In either case, if you have a catastrophic
machine failure, it is entirely possible that your drive groups can
be corrupted or lost entirely.

On all of my critical servers, I use external RAID boxes (generally DEC/
Compaq/HP stuff, but we're an old DEC shop).  Most of our stuff now is
on an EMA12000.  6 7-disk shelves on 6 SCSI busses, 2 dual-ported RAID
controllers with failover, two fibre channel switches, and two
fibre channel cards in each server for multipath failover.  This is
WAY overkill for a simple asterisk server, but there are many smaller
solutions.

One that I've used (that I think is out of production now, but there are
replacements for it) is the Compaq Storageworks RA3000.  It's a little
external drive tower - 7 (If I recall correctly) drives, plus an option
to add an expansion module for another 7, single or dual RAID
controllers with failover, dual mirrored cache, etc.  Configure it, and
hook it up to a standard USCSI controller, and it looks just like
one or more regular SCSI drives.

The advantage to external units being that if something catastrophic
happens on the server, I can simply put a SCSI card (or a fibre card,
in the case of the EMA12000) in another machine, connect it to the
external array, and go.  I can lose any piece of the external system
and still be functional.  Also, since it's a standard interface, I
don't even need to have an identical interface card handy - any
compatible SCSI/FC interface will do.

-- 
Matt White                          whitem@arts.usask.ca
Arts and Science Computer Labs      University of Saskatchewan

It sure is Monday... Ain't it a sin
I've gotta work my way thru the week again.
	- Mark Chesnutt..."Sure Is Monday"

On Fri, 2004-01-16 at 16:55, Robert L Mathews wrote:
> At 1/16/04 7:25 AM, Andrew Kohlsmith
<akohlsmith-asterisk@benshaw.com>
> wrote:
> 
> >That's pure bullshit -- I use software RAID *specifically* because
I value
> >my data.  I don't want to buy two hardaware RAID controllers to
have one
> >sit on the shelf just in case the first dies... and if the second dies 
> >you're SOL because they've lasted long enough that they're
no longer
> >available.  Linux software RAID is available on any Linux system and if
the
> >system blows up I can put the drives in another system and *not* worry 
> >about it not being detected.
> 
> Yeah, I couldn't agree more.
> 
> We originally thought hardware RAID was the way to go, and we bought a 
> couple of fully loaded Dell PowerEdge 2550s with SCSI hardware RAID 5 
> arrays at about $4500 a pop. We also bought a PowerEdge 600SC for around 
> $900 with lots of disk space to use as a network backup machine (backing 
> up the 2550s) with Linux software RAID 5. I've also had a crappy old 
> desktop machine running Linux software RAID 1 for a couple of years.
> 
> It turns out that the software RAID is just as reliable (more so, in fact 
> -- we have had a number of lockups on the 2550s that appear to be due to 
> the hardware RAID subsystem locking up, and the software RAID machines 
> have never done that, even though the backup server does more disk I/O 
> than the others). The software RAID on the 600SC is faster than the 
> hardware RAID in bonnie tests.
I believe there is a recall option on those machines. So far no one has
identified what exactly is the problem there. I was reading the aac-raid
list for a while, some people point the finger at the firmware on the
disks, and some at the drivers. Either way, there are a few machines
that Dell acknoledges trouble with.