[NTLUG:Discuss] Re: High end PCIe storage/RAID adapters -- no PCI-X slots, but only PCI-X storage

[Bryan J. Smith]

Chris Cox <cjcox at acm.org> wrote:
> It's a tunnel (from everything I've read).  That's why
> many say it's fixable after release on some
> implementations.

I'm sure it's tunneling when you connect to a SAS hub or
other intermedia controller, and the end-device has SATA
disks (well away from the original host adapter).  That's
asking for a lot of headaches and I believe you when you say
there's much work to be done -- i.e., "fixable after
release."

But from what I've seen (and I could be wrong), at the local
host adapter, it's just 4 wires right into the PHY/ASIC.  To
get SATA, all the ASIC does it drop the SCSI-2 command set
and drive it directly.  That's why most host adapters can
"get SATA for free."  They aren't doing SAS, but SATA itself,
and the host adapter just handles any queuing/commanding
(just like today's intelligent SATA controllers).

> True... you definitely want the Xscale chip for good
> RAID5.  I've heard the 2E hasn't lived up the it's
> slightly older brother, the 2X.

Yep, because the IOP332 is actually an IOP331 with a PCI-X to
PCIe bridge.

The BCM8603 is an IC with both PCI-X and PCIe, and can be a
bridge between the two if needbe.

> Good to know.  Didn't know they had something ready for
> SAS.  Maybe SAS will become a reality shortly.

Well, I'm sure there will be a lot of headaches early on. 
Especially as vendors implement the SCSI-2 command set over
SAS differently, as well as the hubs, multi-targeting, etc...

But it looks like every SAS host adapter -- at least the ones
with RAID intelligence on-board (i.e., an
ASIC/microcontroller) -- can handle SATA drives directly as
well.


-- 
Bryan J. Smith                | Sent from Yahoo Mail
mailto:b.j.smith at ieee.org     |  (please excuse any
http://thebs413.blogspot.com/ |   missing headers)