[KLUG Hardware] Re: SCSI card recommendation -- _know_ your technology (you obviously don't)

[Bryan J. Smith]

--=-ZYdmCDphTVXceQw1NbS8
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Tue, 2002-11-05 at 10:45, Adam Williams wrote:
> IBM IPS. (ServeRAID III)

Where are you guys finding these so cheap?

> Those IDE RAID SCSI-Emulators sell for what?  > $200.

No "offense," but I _do_ take "offense" when people do _not_ understand
the technology involved.  Don't be _ignorant_.

First off, they are _not_ "emulators."  Don't confuse 3Ware cards with
those "BIOS-only" RAID cards like the Promise _Fast_Trak and many other,
cheap add-in cards and on-mainboard ATA RAID controllers.

Secondly, they _do_ make _exact_equivalents_ of ATA RAID cards to SCSI
RAID cards.  The Adaptec 2400A and the Promise _Super_Trak (not the
"FastTrak") use the same damn i960+SDRAM design of 90% of the SCSI RAID
controllers out there.  In those cases, you are getting the _same_
design, only ATA controllers are used for storage, instead of SCSI.

Third, 3Ware cards are non-blocking I/O ASIC+SRAM "storage switches."=20
They are the absolute best performing cards for RAID-0, 1 and 10 (aka
0+1) -- period.  The blocking I/O nature of SCSI, especially when
combined with an on-board microcontroller (uC) and latecy ridden DRAM
technologies, is a disservice.

Comparing ASIC+SRAM to uC+DRAM is like comparing dedicated network
equipment to a PC-based switch/router.  The former is faster and has
reduced latency whereas the latter is more flexible, but introduces a
great amount of latency.  When you're on a LAN, the former makes more
sense because you're often talking direct and doing little buffering,
whereas the latter makes for sense for routing, WAN connectivity and
Internet caching when buffering and scheduling is more important.  Same
ordeal with storage, the former is best for direct I/O operations like
RAID-0, 1 and 10, whereas the latter is better for more advantage
operations like massively random RAID-5 writes (e.g., databases).

> To me that just seems silly.

Since people like yourself obviously don't stop to investigate _why_
some things are better for X and other things are better for Y, then
yes, it all seems silly.

Technology is best implemented where appropriate.  There is no holy
grail of "one size/implementation fits all."

> People whine about using IDE instead of SCSI for the cost?!  One can
> get a great SCSI controller for ~$70 and the highest end drives
> (15,000RPM big cache IBM) of 18Gb can be had for ~$300.

18GB for $300 might be good if you don't need much storage.  But for
file servers that need hundreds of GBs, that adds up quickly.

> Lower end drives can be found much cheaper.

180GB of ATA is now $300.

> And I'll still pit a good SCSI box against IDE any day.

Depends on your need.  ATA drives are "dumb" block I/O, SCSI is
intelligent, scheduled I/O.  ATA before UltraATA didn't offer DMA with
error-checking, but UltraDMA and later does, bringing "equal" to SCSI
for desktop duties.  I didn't start adopting ATA on the desktop _until_
UltraATA came out with CRC checking that made it "as good" as SCSI with
parity.

Add in the non-blocking I/O nature, and ATA is much better for many
applications.

Then move onto RAID.  Now the "brains" is in the controller design, not
the storage bus.

Do you need RAID-0, 1, 10, have lots of RAID-5 reads or sequential
writes?  Then 3Ware's non-blocking I/O is the best option.  If you have
lots of RAID-5 random writes, like for databases and some web
applications, then a quality SCSI RAID card would be better.

> Lets take a crappy technology, because its cheaper,=20

ATA serves its design and purpose -- a "dumb".  SerialATA improves many
things by finally removing that _stupid_ "slave" option (which I _never_
use).

> sink a pile of money into it so that it *resembles* a real technology?=20

Again, ATA serves its purpose -- and better than SCSI in many cases.

> How does that make sense?

No, because you didn't give specifics.  You made analysis based on your
[ignorant] assumptions.

-- Bryan

P.S.  If "ignorant" is too "offensive" of a word, Find/Replace it with
"unfamiliar" instead.  ;-P

--=20
Bryan J. Smith, E.I.            Contact Info:  http://thebs.org
A+/i-Net+/Linux+/Network+/Server+ CCNA CIWA CNA SCSA/SCWSE/SCNA
---------------------------------------------------------------
The more government chooses for you, the less freedom you have.

--=-ZYdmCDphTVXceQw1NbS8
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQA9x+5GDjEszaVrzmQRAuQfAKClPq+8pv5ngdeGzfvji4B1WxSlPQCeOCm/
mRtjzQspdkVbSsxIn4ZrQpY=
=Ximu
-----END PGP SIGNATURE-----

--=-ZYdmCDphTVXceQw1NbS8--