new tech rave.... Was: rsync and fP - never mind. *sigh*

[Bill Vermillion]

When asked his whereabouts on Sun, Oct 31 22:08 , GCC Consulting 
took the fifth, drank it, and then slurred:                      


> > [mailto:filepro-list-bounces at lists.celestial.com] On Behalf 
> > Of John Esak
> > Sent: Saturday, October 30, 2004 2:28 AM
> > To: Fplist (E-mail)
> > Subject: OT: new tech rave.... Was: rsync and fP - never mind. *sigh*

> > > Now I have to demand more disk.  :)

> > Yeesh, disk space is SO cheap these days... ask for 140GB!  
> > It's unbelievable.  Hell, I just got 2GB on a usb jump drive 
> > the size of a big peanut for about $400. Of course, a 
> > non-moving media device is too slow for most things we do, 
> > but it is sure good enough to carry around whole filePro 
> > systems and applications.  As for the latest 140GB and higher 
> > SCSI drives...

...

> Last weeks issue of e-week discussed the cost of storage. And
> although the price has been coming down, the physical limits
> of magnetic disk storage is being reached. They discussed
> holographic storage which about to come out. IBM also has some
> new technology in the offing. However it is very expensive, for
> now.

There was talk we were reaching the paramagnetic limit [the maximum
amount of changes that can be stored in a given area] but last year
developments at IBM showed that limit was probably 20 times smaller
than previously thought so that we probably have at least another
decade before we run out of room with current technology.

> As for why that small computer has only 256MB, it may be a
> power and cooling thing. More memory=more heat+more power. I
> have a number of clients running XP Pro on 256MB. The key here
> is all they run is FP 100% of the time. The machines which need
> to do more have 512MB or 1GB of memory.

Fast CPU's suck far more power than RAM.  On reason Intel canceled
their 4GHz CPUs was just that.  Instead they opted to make the
current designs more efficient.   What a novel concept for the
computer industry.  A P4 will take about 250W of power.  A high
end video card can take another 75W.

Remember the original PC with it's 60W power supply?

> I have found that fp responds to brute horsepower best.
> Adequate memory for the OS and a speedy processor alone with
> good i/o subsystems makes for happy clients. The cost of this
> keeps coming down.

It's I/O subsystems that keep getting beat up upon.

Storage is getting bigger and cheaper.

Apple, who has never been now for low prices, is selling
a 5.6 Terabyte RAID array - with 2Gb fiber channel connections for
at least a 400MB second data transfer rate - for $12,995.

That's a 3RU that is essentially plug and play.

The 400GB external drives that are on the market now are running
from about $375 up.

The problem is that users want/need more and more speed as
the applications grow, and more ways of handling data are found.

Just two years ago Sears re-evaluated their data and found they
could mine it more efficiently to increase sales to it meant
their 7 Terabyte storage requirements would increase to in
excess of 70 Terabytes.

But the problems we have are the I/O devices.

SCSI has always been the performer, but for many the IDE and EIDE
have been good if the OS doesn't need to do a lot of multi-tasking.

But we are limited by the interfaces there.  You may have noticed
that the top IDE speed is the UDMA-6, aka UDMA-133.

Current drives are approaching that interface limit.  The fastest
of the ATA/SATA drives I've seen are the the little Samsung
units being able to transfer data from the disk to the internal
buffer at 80MB/sec.  That's not too far away from the ATA-100 in
most devices.

SATA - at 150Mhz - not that much faster than the 133Mhz UDMA and
only 50% faster than the 100MHz - won't be a great step as the
internal speed of the drives increase.

The PCI bus runs at 33MHz and is 32 bits for for 132MB/sec transfer
rate.  The data transfer rate internally on my P4 2.4GHZ with
dual DDR chips on a 400MHZ FSB is close to 8GB/sec, so you can
see much performance of today's machines is limited by
the bus structures.

PCI-66 moves the bus twice as fast and there for doubles the data
rate.  And PCI-X is 64 Bits wide at 133MHz, so it's about 8 times
as fast as the standard 32 bit wide 33Mhz PCI.

Don't expect anything to replace rotating magnetic media for
density and price in the foreseeable future. 

A paragraph heading in this months issue of Infostor reads 
"Todays' 2 Gbps Fibre Channel technology is sufficient for most SAN
applications".

Note the key word most.  Describing current video needs  current
HDTV transfer rates are 165MB/sec [approx 1.3Gb/sec].  The next
generation of HD acquisition devices will be at 240MB/sec
which means a move to 4Gb fibre channel.

What we have today would have seemed like fantasy just 10 years
ago. Who knows what the future holds.

Bill
-- 
Bill Vermillion - bv @ wjv . com