On January 18, 1996, Evan Rudderow posted:
>In the January '96 issue of The 3000 News Wire Ron Seybold recounts Ken
>Paul's visit to DMD regarding the reliability of 2GB disc drives, at one
>point Ron writes:
>> Paul was following up on his notes and meetings as we went to press, but
it's clear
>> the existing failure rate of the 2GB devices is here to stay. HP has
stopped
>> making the drives, but some are still being stocked by the Computer
Systems
>> Organization for use in HP3000s. Boise engineers have long since moved on
to
>> designing 8.75 GB drives, while the Disk Memory Division is now
manufacturing
>> new 4GB units just released to HP3000 customers last month.
>Does this mean that 2GB drives will be dying a slow death -- or that they
>will not be obtainable much longer? Frankly, principally because of I/O
>throughput considerations, 2GB is about as much as I'm willing to put on a
>spindle.
>So, Ron -- or anyone else -- what's the future hold for us when it comes to
>disc drive capacities?
Sorry for the delay in responding. This digest never made it to me so I had
to pull
it off the index and I've been working on my treatise/tome on this subject
for
InterexPress (I will also post it to HP3000-L for the masochists).
Evan's first question about dying a slow death or not be obtainable much
longer:
I'm not sure when they will be dropped from the HP price list but I do know
that
DMD is no longer making these drives (according to what was said in Boise).
These drives are still stockpiled at CSO however and I do not know the
nuclear
half life of this stockpile. I'm sure you will be able to get remanufactured
ones
and ones from resellers in the forseeable future just like you can pick up
"Eagle"
drives still if you really want one.
The discussion never got to I/O throughput considerations of the new 4GB
drives
compared to the 2GB drives or the newer 8.75GB drives versus either. I, like
Evan,
have never liked people consolidating their systems on fewer and fewer drives
with
greater capacity because of potential performance problems. Disc I/O has
always
been the one thing that was not growing at the meteoric pace as things like
CPU
speed and power. So to me consolidating has always been like having more
people
share fewer pipes.
I hope this also answers the question about disc drive capacities although I
think
we (HP3000) may see the 8.75GB drive before next year.
Later the topic changed to RAID and AutoRAID:
Ron Seybold said:
>Ken Paul also said that AutoRAID is very big in HP's futures for
>3000-level storage devices. If one of these fault-tolerant RAID units is
>made of 10 2Gb-devices (or hopefully, more reliable ones), is it a 2Gb
>unit anymore? Does the redundancy improve the value beyond 2Gb reliability
>rates, and at what kind of cost?
To which Jeff Kell responded:
>Similarly for RAID arrays - if the controller interface is single-threaded,
>that may not be a Good Thing(tm). Of course it will sell memory, to hold
>all those prefetches to keep I/O rates to a reasonable level.
And Evan Rudderow responded:
>For me, the jury is still out on RAID with respect to performance. When I
>first started reading about RAID a few years ago it occurred to me that,
>while RAID would protect data integrity, it would do so at the cost of
>performance -- because while in a non-RAID environment 8GB might have been
>on 4 spindles (each with a disc controller) and two channels, post-RAID it
>would be 5 spindles (with one controller for all five mechs) on one channel:
>a reduction of, say 120 aggregate I/O's per second down to 30 I/O's per
>second. Some of my friends didn't see that connection, converted to RAID 5
>arrays and subsequently got to deal with performance problems. In fact,
>I'm still hearing war stories about sites that went RAID and saw their disc
>performance fall through the floor -- I recall that there was a posting on
>3000-L within the past week or two...
While visiting the Storage Systems Division (SSD), I was introduced to their
"ICE"
product which was announced February 1995. It is also known as the XLR1200
Advanced Disk Array with HP AutoRAID Technology. This product won the
Editor's
Choice Award in the December 1995 issue of Unix Review. I have copies of a
five
page marketing white paper, a technical data sheet for OEMs and reprints of
the
Unix Review article.
According to the Technical Data Sheet, a 19 inch enclosure can handle 12
1-inch
units or 6 1.6-inch units for a total capacity of 24GB. Six of these can go
into a
rackmount enclosure with 144GB total capacity. These capacities could
obviously
change with the 8.75GB drives in the future. Each unit has two power
supplies,
three fans and two controllers which are totally hot-swappable. The discs
are also
hot-swappable. I was able to see one of these units and they are indeed
impressive.
The data sheet also mentions 16MB standard RAM - expandable to 24MB.
This new technology is not the RAID that Evan is talking about above. RAID
devices
are very difficult to configure and a choice must be made of what RAID level
to use.
These new units are "automagically" configured and track the usage of your
data.
Your most active data will be kept in RAID 1 format for best performance and
your
less active data will be in RAID 5 format. According to my notes, 10% of the
device
will always be in RAID 1 format and the data is checked at the block level of
64K
blocks. You also have an Active Hot Spare but someone else may have a better
explanation of what this means than I can at the moment. To use all of the
buzz
words moving the data between RAID 1 and RAID 5 "automagically" is called
Dynamic Data Migration (DDM not to be confised with the DMD in Boise).
No official word on when this will be available for the 3000. Does the word
"soon"
appeal to anyone. I could see these devices also being available before the
end
of 1996 but I could also be wrong.
Hopefully this gives everyone more to think about. I never said that I would
give
definite answers to anything for I am just a spewer of (hopefully correct)
information.
Respectfully submitted for debate,
Ken Paul a.k.a. [log in to unmask]
|
