[Info-vax] OT: Steve Wozniak

[mjmahon]

On Aug 17, 6:35�pm, Arne Vajh�j <a... at vajhoej.dk> wrote:
> Neil Rieck wrote:
> > I, along with 700 others, just had breakfast with Steve Wozniac at a
> > conference sponsored by Waterloo Ontario companies like RIM, Dalsa and
> > Open Text.
> >http://www.communitech.ca/en/
>
> > Boy, I thought I was an optimist but this guy's optimism is
> > overflowing and infectious. Why would you OpenVMS people care about
> > this? The Woz now works for a company called "fusion i/o"
> >http://www.fusionio.com/
> > and he mentioned that big companies, like HP and IBM, are using
> > "fusion i/o" solid-state storage technology to set new TPC transaction
> > records while beating fiber-connected storage arrays by 10 times or
> > more (and costing a whole lot less)
>
> >http://www.computerworld.com/s/article/9100218/HP_adding_solid_state_...
>
> > In their view, multi-core CPU processors will only get faster which
> > means that magnetic storage will continue to starve them of data. They
> > feel that solid-state storage will become the primary system memory
> > while hard disks will be relegated to doing off-system backups.
>
> > Now let's see if their vision comes to fruition.
>
> SSD is a lot faster than rotating disks. And will undoubtedly
> make a TPC benchmark scream.
>
> But SSD is a relative unproven technology regarding reliability.
>
> The first generations of SSD (the first generations of the
> technology that is used for SSD today - SSD is a very old
> concept) were not very reliable for many writes.
>
> Intel claims that their newest SSD's are as reliable as
> server disks.
>
> I think a lot of system managers will want to get some
> practical experience before putting the companys
> core systems on SSD.

It is unlikely that the SSD will be used without redundancy.

If mirroring or other RAID redundancy is used, the probablilty of
losing data as a result of chip failures is reduced essentially to
zero.

Wear leveling is very effective in spreading the "write" load across
all blocks of an SSD, and failure is not sudden, but is a gradually
rising BER rate, while errors remain fully correctible by the built-in
error-correcting codes.  This allows easy "chip sparing"
strategies, so that even a single SSD array need never manifest a data
error resulting from endurance issues.

It has been long realized that a journaled file system or database can
be updated as fast as the log can be updated.  SSD is a technology
well-suited as a non-volatile log storage medium, and should allow the
primary disk storage to appear to update at the speed of the SSD.

Disk reads can also be accelerated by caching, but for random
requests, caching is not very effective in reducing latency.  The huge
numbers of "heads" offered by a large array of NAND flash, combined
with the very low intrinsic latency, could offer the order of
magnitude access speed increases referred to by Woz, but only if the
SSD is the primary storage medium for the database--which will require
significant price reductions for most applications.

Right now, flash is about 10x the speed and 10x the cost of magnetic
storage, so if the speed is worth it, flash SSD is the way to go.  For
most people, using a much smaller amount of SSD as a cache for
repeatedly read data and as a log for updates will get them a lot of
performance bang (1.4x-2x) for relatively few bucks--a big win.

-michael