Highend system... Will it fly? Any issues?

[kony]

I thought it may be 'Mean Time *To* Failure', as I said, I know what MTBF
means.

I last repaired a failed HDD a few months back, by swapping
controller-boards/PCB's with a similar drive, just until I got the data off
it.

I am fully aware that mechanical failure of a drive is irrepairable.
(Without the proper equipment, and even then only for data recovery).

Thanks.

MTBF is pretty much irrelevant, given that all brand-name
manufacturers now have reasonably reliable product... We just
have to assume they have an accurate figure, since even those
"defective" drives from past years didn't have correspondingly
low MTBF figures.

Anyone that's been around for a few years has data that's worth
far more than the cost of a single HDD.

Even as interesting as Seagate's new 5-year warranty is, I'd be
much more comfortable knowing what, if anything, leads them to a
belief that their drives will last longer (on average, compared
to competitors warranties). I'd rather have a drive that lasted
20% longer, with no warranty, that a drive failing sooner with an
assurance that it'd be replaced.

Thease days, storage is 50 cent per GB, but the data is worth
more than that.

 

[kony]

Building a rather expensive system based on a dual Opteron mainboard and
want to know if there are any issues I should expect...

Mainboard - Tyan Thunder K8W S2885
CPUs - (2) Opteron 242 retail
Memory - (4) 256meg DDR400 OCZ Registered ECC
Video - ATI Radeon 9800 All-in-Wonder PRO 128meg 8xAGP TVOut DVI
RAID card - Highpoint 1820A PCI-X 64bit 8port SATA RAID card
HDD - (1) 34gig WD Raptor 10,000 RPM SATA drive (for OS)
HDD - (9) 250gig Maxtor 7Y250M0 Sata drives (RAID 5 for data)
DVDRW - BenQ DW822A DVDRW
Floppy - 3.5" floppy
Case - Lian-Li V2000
PSU - Antec 550watt 24pin EATX power supply
Fans - 2x80mm 2x120mm

...any issues the the above configuration?

...also, does this mainboard REQUIRE Registered ECC memory? I've had
conflicting information about this.

Any comments are welcome.

I wouldn't use 4 x 256MB memory. Four modules just to reach 1 GB
on a machine of that class is a waste of slots. Multiple memory
manufacturers have shifted production, i expect that if you "shop
around", you may find good deals on good memory in the near
future... at least till the western world's school buying-season
starts again. In general memory market has dropped, except
spot-purchase of high-end, DDR500+ spec.

The power supply may work, given the card's ability to control
spin-up, but I still wouldn't trust that much data, components,
to a single PSU. I don't yet have any systems at home with that
many drives, but my main rig already has two PSU, and a network
storage system enough to backup that and more... in other words,
no single-point unrecoverable failure could prevent access to
any, ANY data. IMHO, HDDs, are a necessary evil, but today's
storage capacity is nowhere near worth value of data stored. Too
much data stored on a single point is a recipe for disaster.

In other words, before I'd spend as much on a single-node
storage, i'd have NAS that can handle at least as much... maybe
you do, it was just a comment.

 

[CBFalconer]

.... snip ...

Anyone that's been around for a few years has data that's worth
far more than the cost of a single HDD.

Even as interesting as Seagate's new 5-year warranty is, I'd be
much more comfortable knowing what, if anything, leads them to a
belief that their drives will last longer (on average, compared
to competitors warranties). I'd rather have a drive that lasted
20% longer, with no warranty, that a drive failing sooner with an
assurance that it'd be replaced.

Thease days, storage is 50 cent per GB, but the data is worth
more than that.

There are two steps to data protection. One is protection against
internal destruction by use of reliable memory, which can only
mean ECC. Without that (or, a distant second best is parity) the
internal data paths are not trustworthy, and there is no mechanism
for even reliably detecting a fault.

The second is backup. The quickest and thus most likely to be
used is a second identical drive, which holds a mirror of the
primary drive. This can usually perform a complete backup in a
matter of minutes, and if the primary fails can step in to replace
it. Note that the backup must still be performed at discrete
intervals, otherwise any faults can immediately destroy the
backups value. Luckily this is all quite feasible today with the
low prices of HD storage.

After that such things as networks, removable drives, tapes, CDs
and DVDs can provide off-site disaster protection. Removable
drives fit into the above backup mechanism, and are probably the
most likely to be effective. CDs and DVDs can provide long term
backup.

At least as I see it.

 

[Noozer]

Building a rather expensive system based on a dual Opteron mainboard and

I wouldn't use 4 x 256MB memory.

The reason for this was to keep two sticks on each CPU, since each Opteron
CPU has it's own memory bus and to allow for interleaving. There are eight
memory slots on the board so there's still lots of room to upgrade. Also,
(and I haven't verified this), depending on how the CPU's are loaded, we may
find better performance putting 3 sticks on one CPU and 1 stick on the
other. (Something I want to test out)

The power supply may work, given the card's ability to control
spin-up, but I still wouldn't trust that much data, components,
to a single PSU.

The was done based on the cost of a "dual PSU" ready case. Prices went up a
great deal when we found a case that could handle this many drives AND dual
PSUs. There is room to mod this chassis for a second PSU if we need it, but
we'll see what happens with this first.

....Thanks for all the comments guys! Definately good stuff to think about.