[KLUG Members] Re: New machine - Linux hardware?

[Bryan J. Smith]

Adam Williams wrote:
> I was referring to the case.  I lust after one of those two
> power supply fifteen drive bay double wide monster cased.
> Too bad they cost more than the rest of my computer combined.

Even the PS on the PC Power&Cooling boxes are excellent.  Their 350W
does just under 215W on 3.3+5V, and the 425W does just under 300W. 
Again, that's excellent!

> I've looked at IDE Raid devices and for the same controller
> cost I can get a very nice SCSI Raid controller.

Agreed.  But why is this?  It is obviously because _real_ SCSI RAID
controllers are more numerous, i.e. economies of scale.  The number
of _real_ IDE RAID controllers are few and far-between, limited to a
handful of companies (mainly 3Ware and Adaptec).  But the opposite
happens when you get to the storage devices themselves, where IDE
"makes up" for its cost different.  That's why I recommend IDE RAID
for raw storage requirements for SOHO networks.

> SCSI has other advantages as well:
> more devices per bus, more types of devices, and external
> devices and cabinets.

Of course!

> IBM ServerRAID III triple-channel PCI cards with PowerPC
> processor and cache memory sell on E-bay for less than $100,
> and they FLY!!!!

Er, "refurbished/used" hardware is not a good argument, because
these _real_ IDE RAID controllers haven't been around long enough. 
But it does go back to the "economies of scale" argument, where SCSI
RAID controllers are just more ubiquitous.

> On E-Bay I got an eight bay drive tower with fans and redundant power
> supplies for ~$100.  This is great for both expanded capacity and heat
> management (the drives aren't in the case, at least not all of them
> anyway).

That's quite a deal!  Again, it's difficult to get refurbished IDE
drives and, since they are so new, RAID controllers.

> Also (this is ancedotal), in my experience, my IDE drives have a
> drastically shorter average life span than their SCSI equivalent,
> althought they are supposed to be the same physical drive unit.

Of course.  They are less tested, pack more density in the same
size/platter (reduced reliability), don't have sector remapping,
cannot be easily low-level formatted (which goes a long way to
"refurbishing" SCSI drives regularly).

And it is actually _rare_ that the IDE and SCSI drives are the same,
physical unit.  Quantum tried doing this with little success, and
Seagate does not.  I haven't seen WDC doing it as of late, and
Maxtor is an IDE-only distributor.  I don't think Fujitsu ever made
their IDE-SCSI products the same mechanics either.

> In my experience a power user with a powerful box can reach this point
> quite easily.  But it all depends on what you do.  If your a compiling,
> number crunching, SQL query-ing fool then IDE, to put it bluntly,
> stinks.

Of course!  Threading is not IDE's "strong suite" by any means.  But
if you use a _real_ ATA-RAID controller, that looks like a SCSI
target from the OS' standpoint, much of that is hidden.  Heck, some
of them can even do sector reservation and remapping for IDE
drives!  But when it comes right down to heavy usage, SCSI is going
to be hard to replace.

> I've had multiple (I loose count) transactions with both
> companies and had no problems with either.  But obviously
> YMMV.

I still deal with both.  HiTechCafe hasn't screwed up on my last 5
orders, so they are "in the clear."  The CompGeeks issue was due to
ordering lots of parts, which held the orders up, due to
unavailability of 1 or 2 of the parts, and were eventually
cancelled.  CompGeeks seems to have never head of "backordering."  I
now make my orders with smaller numbers of items.

-- TheBS

-- 
Bryan J. Smith   mailto:b.j.smith@ieee.org   chat:thebs413
SmithConcepts, Inc.           http://www.SmithConcepts.com
==========================================================
Linux 'Worms' exploit known security holes that were fixed
3-12 months earlier.  NT/2000 'Worms' exploit unknown se-
curity holes that won't be fixed for another 3-12 months.