Torben
Thanks for your detailed response. Furthermore, before people start
accusing you of being an ADPT raging bull, as far as I understood
your posts, you seem to be a powerful supporter of their technology,
and have offered far fewer views on the quality of their stock !
FWIW, I will offer my "answers" to those earlier questions, as I
am desperately trying to understand how much mileage is left in the
SCSI market. I suspect from your post that we see things somewhat
similarly......
Q1. What will be the best HDD interfaces in the near future for -
a) inside the box where cost is an issue ?
Ultra-IDE, Ultra-DMA, 66, or whatever it's called this week. Basically
it will be the highest performance IDE derivative that can be created.
The majority of these controllers will be part of the Intel core logic
though some PC and low-end workstation vendors will use 3rd party
controllers. HDDs will become increasingly cheap and higher bandwidth
and volumes will mean there's no low cost alternatives. In addition,
improvements in the driver software will be able to replicate the
re-ordering algorithms that have traditionally been the domain of the
"transaction" based SCSI bus, so that the end result will be low cost
AND high performance.
b) inside the box where cost is not an issue ?
You *might* still use SCSI, but why ? At a similar price to one
high performance SCSI drive you could have two fast IDE drives, and
provided the driver software is good enough you'd get better
performance and higher capacity for no extra cost.
c) inside the box if increased configurability is required ?
P1394. It's fast, and could easily become the "device bay" HDD I/F
of choice. Sure it's not going to be as cheap as IDE, but it does
at least offer longer cable runs.
d) outside the box where cost is an issue ?
Today SCSI (because it supports long cable runs and there's no other
choice. SSA ? I don't think so).
HOWEVER, I think 1394 could change this. IF 1394 becomes the
device bay HDD I/F, then the next logical development is to have
"device bay carriers" outside of the box. In other words you can
add additional device bays outside the box, so that you can plug in
all those extra device bay devices. 1394 would make it easy to join
the dots.
e) outside the box where cost is not an issue ?
Fibre channel - it has a performance and distance advantage.
Q2. Which of the above will be the largest markets ?
Short term - Ultra-IDE. Longer term - 1394 ?
I think the thing that really worries me about SCSI is that it's
traditional domain - disk expansion outside the box - is rapidly
being superceded.
Ultra IDE drives are getting big, fast and are cheap. Most computers
have more than enough space and spare HDD channels to add these inside
the box, so the "expansion" market is drying up.
Secondly, where you have to have large disk arrays outside the box,
you need faster interfaces that support longer inter-connects. Again
SCSI won't hack it.
The final nail in the coffin, is that even the workstation vendors
are beginning to drop SCSI as a standard expansion interface. Sun
is probably the highest unit volume shipper in this market, and their
latest volume offering (the Ultra 5 & 10) do not have SCSI as
standard - for the first time ever. Sure these machines have PCI
card slots (also a first - for Sun !) which means that you could
easily add a SCSI card, but I don't see this as much of an endorsement
for SCSI. Surely most SUN users will just add IDE HDDs inside the
box, and when they need to connect up a SCSI HDD (legacy reasons ?)
won't they just join these to an older machine and rmount them on
the increasingly fast networks ?
.
I think it would be premature to say that SCSI is dead, but I think
the above *might* be a reasonably cohesive argument for why it might
quickly go away.
One lateral thought that nobody has proposed - ADPT aren't the only
SCSI controller vendor. I *think* that QLogic are also quite big
in this market - Does anyone know what's happening to their sales ???
Mark |
