[Beowulf] Ethernet connected drives
landman at scalableinformatics.com
Thu May 8 13:12:55 PDT 2014
On 05/08/2014 03:36 PM, Lux, Jim (337C) wrote:
> What these will be good at (and what the Active Disk research
> espouses) is applying very simple filters (e.g., simple greps) to
> avoid pushing data across buses needlessly. Basically, only the data
> you want to come out that you will then compute with will be
> returned. The compute- and ram-intensive tasks will still be
> reserved for the real machine.
> Decent paper on one approach to this using SSDs (since they already
> have compute and memory on-board to deal with all their firmware
Actually the compute available on SSDs is pretty limited, and for a
number of reasons, this won't be changing much any time soon. The paper
notwithstanding, most of the folks we've spoken to are taking a waiting
position on this at best.
FWIW: this tight coupling between computing and storage is something
advocated (strongly) by a number of groups and companies (including
me/us). Its simply too compelling from a performance standpoint over
the long term, as data motion only gets harder with data volume.
This said, the limited processing capability on a drive may be able to
run a TCP/IP stack at near GbE speed. I doubt it could do much more
than that without driving the drive costs up. Put a 10GbE chipset on
each drive, with signaling and other bits, and you'll be adding a few
hundred $$ to each part. 1GbE won't add much, but there will be
precious little processing power left over for applications. Not to
mention very limited ram. The other aspect of this is the ram on the
units would likely need to be flash backed, and not just the cache ram
in this case.
All of this drives up power and cost. What is the benefit? You get
additional processing power, on par with Calxeda ARM chips (think 16 or
32 bit CPUs), with all the drawbacks, less ram, more complexity in code.
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web : http://scalableinformatics.com
twtr : @scalableinfo
phone: +1 734 786 8423 x121
cell : +1 734 612 4615
More information about the Beowulf