[p4] NFS mounted depot

[Weintraub, David]

From: Paul Goffin [mailto:Paul.Goffin at aepnetworks.com] 
>> I had a long talk about this issue with the Perforce support. Most  
>> corporate Unix environments don't backup local drives and normally  
>> don't even give you access to the local disks. Everything is NFS  
>> mounted except for the OS, so if what Perforce states is true, you  
>> couldn't use Perforce in many corporate environments.

> Of course you can.  You run Perforce on your servers, not on the
client machines.

> What you seem to be saying is that your IT department doesn't
recognise
> Perforce as a mission critical process and your SCM data as mission
critical.

> If that's the case, your problem isn't with Perforce, it's with your
> IT management who need taking out and shooting.

As a person who once worked in an IT department, I remember being on the
other side of the battle line. Why can't you people just get us more
disk space? Why is diskspace so expensive? I can by a 100 gigabyte disk
from CompUSA for $110 and you guys want to charge me 10 times that rate?
I need to run my application on an X type of drive. Why can't you guys
get your act together and give that to me? Where am I suppose to store
my MP3s I downloaded?

Working the backend wasn't so simple. We had thousands of racks of
disks. We had disk failures and server failures happen almost every day.
However, 99% of the failures went unnoticed by our clients because we
had as part of our mandate that we strived for 100% uptime. We had
snapshots. we had disk mirroring. we had RAID drives. I am pretty
certain that we even had technology which came from another planet with
a highly advanced civilization because it was able to do things I
thought were absolutely impossible until I worked in IT.

Each and every disk had to be backed up at least once per day --
sometimes more than once. And, we had special requests coming all over
the place. Databases had to be locked before being backed up. Same with
ClearCase VOBs. Yet, almost everyone needed their read/write access
throughout most of the day and even 24 hours per day since we had
offices around the world which meant we could never lock the disks for
backups.

We also had generic servers waiting in the wings that could be pressed
into service whenever any of our other servers failed. To us, local
disks on machines were a big fat pain. It meant that we couldn't have a
generic server acting as backup since that generic server wouldn't have
the same data as the server it was replacing. It meant that backups were
much trickier, specialized, and untested in a crisis.

Even if we could somehow figure out a way to do an unobtrusive backup,
what would our response time be when a disk failure or server failure
occur? Say we are supporting 100 technical people on a particular
server. Say it costs our company about $300 per person (including
salary, benefits, furniture, air conditioning, office space, etc.). Say
we could bring everything backup in four hours without a the loss of a
single byte of data. That's $120,000 in downtime. That wouldn't be
acceptable to our department or our client. That's why we discouraged
"special setups".

I might not like it when IT tells me that disk space is ten times the
price I think it should be, or that they tell me that my special setup
can't be supported, but I do understand it. I can't even claim that
Perforce is mission critical since there are other version control
systems that will be more than happy to use non-local disks. 

That's why I talked to technical support about the NFS situation while
we were evaluating Perforce. And, that's when they assured me that not
only could Solaris use NFS disks, but that you could store the RCS files
in the depot on NFS drives too.

--
David Weintraub
david.weintraub at bankofamerica.com
david.weintraub at bofasecurities.com