file server for home

[Brian Densmore]

> > I've got to start choosing my words more carefully. I was 
> speaking of
> > software RAID. Software RAID could be done on a single HD 
> (not that I
> > recommend it in any way).
> 
> Whoa there! I'm going to have to give you some shit about 
> that statement. The
> very concept of RAID requires you to have more that one disk. 
> However, you could
> mirror two partitions on a single drive just so that you had 
> two copies of
> something. That way if you were about to change something on 
> a really important
> production machine and wanted a golden parachute to bail you 
> out if in case
> something were to go wrong. Then you would just stop 
> mirroring and perform the
> operation just on one partition, if all goes well remirror. 
> If it doesn't, boot
> off the unmodified partition and your back where you started. 
> However, this
> wouldn't offer any performance advantages. Actually it would 
> reduce performance
> by half since it has to write everything twice.
Exactly, why I said I don't recommend it. It seems nobody is reading the
entire comments here. I said you could. I didn't say you should. In fact
you shouldn't, but you "CAN".

> 
> > SCSI is not that much more expensive than ATAPI.
> 
> Hogwash! You can get an 80GB 7200rpm UltraDMA100 (or 
> UltraDMA133 I think it is
> now) for $160
> 
> The same size for an Ultra160 SCSI version will cost you $500 
> (Going off prices
> from www.pricewatch.com)
37 GB IBM Ultrastar SCSI $159 15K RPM
40 GB Maxtor $85 7200 RPM 

2x speed 2x price. From pricewatch. To me that is comparable prices. Of
course there might be a cheaper model at 7200 (I didn't see any, but I
do have prejudices against certain manufacturers)
Also there may be a widening gap as size increases. I don't really watch
it that closely. So if I seem ignorant please forgive me, but this is
the way I see it. My own SCSI system (not in use right now) is used
primarily for airfoil design (aka 3d modeling) and makes extensive reads
and writes to the disk. So I can notice the speed differences between
7200 RPM and 15K RPM.
> 
> Then lets not forget the controller. A basic Adaptec 19160 
> controller will cost
> you an additional $150 if your motherboard doesn't already have a scsi
> controller built in. (Not many do)
Yes good SCSI controllers will cost you some money. Even bad SCSI
controllers will cost you some money.
> 
> > I said SCSI because of speed [10-15K RPM], up to 2x as fast as
> > ATAPI can get (money is no object argument is no longer 
> valid). And I
> > said recommend, not must have. We are talking about a file 
> server, and I
> > don't know what he is planning on using it for.
> 
> A file server for home.... Home being the key word there. Now 
> even if he does
> have cable or DSL, his bottleneck isn't going to be the hard 
> drive speed. It's
> going to be the internet connection he has. The "money is no 
> object argument" is
> still very much valid. I mean, if you can get your company to 
> foot the bill, go
> ahead and spring for the SCSI. However, unless your going to 
> be doing some
> really freaky shit, you won't notice the difference in a real 
> world environment.
Well, that depends on what you call real freaky sh**. We are talking
about a "file server". So I would guess that it would spend most of it's
time serving up files. Therefore most of the time the hard drives are
being accessed and data transmitted via the network. I didn't see any
reference to internet in there.

<soapbox>
Like I said, that was my recommendation. I stand by it. Although, I
realize that not very many people are going to get a SCSI system for a
budget home file server, because, you can get twice the size for the
same price (even if it is half the speed).

I have 1 SCSI drive myself, it is used for a special purpose machine and
it is absolutely necessary for my purpose. It takes me 30 - 45 minutes
to render a single foil. It takes 1 - 2 hours on an ATAPI disk. Even I
can notice that difference! But then airfoil design is kind of freaky
sh**! I am also working on a rocket engine design (yes I am an amateur
rocket scientist). ;')
</soapbox>

'nuf said,
Brian

P.S. Thanks for the stimulating discussion.