[Beowulf] Project Natick

[Prentice Bisbal]

I heard about this on BBC World News this morning on my way into work. I
waas going to share this here myself this morning.

What isn't clear is how the heat is being transferred from the CPUs to
the seawater. My best guess at the moment is that the capsule's steel
walls conduct heat from the hot air (nitrogen gas, actually), to the
seawater, cooling the ambient "air", and then the air is circulated just
using the chassis fans. It's  possible there are some supplemental fans
to circulate air around the capsule, but based on the photos of the
racks being loaded into the capsule, it doesn't look like there'd be
much room for that. This would mean the processors are still air-cooled
themselves.

Has anybody seen any more details on how the cooling actually occurs
withing the capsule?

Payload
12 racks containing 864 standard Microsoft datacenter servers with
FPGA acceleration and 27.6 petabytes of disk. This Natick datacenter
is as powerful as several thousand high end consumer PCs and has
enough storage for about 5 million movies.

Since they are equipped with FPGAs, it doesn't sound like they're doing
routine workloads, so this technique might not be transferable to you or
me. I'm assuming the FPGAs will get much better performance per watt
than a general processor, reducing the heatload in the capsule vs. doing
the same workload with only x86 processors. Does any one know what the
intended workload of this system is?

https://www.bbc.com/news/technology-44368813
https://natick.research.microsoft.com/
I must admit my first thoughts on hearing an item about this on Radio
Scotland is that now that humans have laid waste to the surface of the
Earth we are going to boil the oceans.
My second thought is for the poor HPC engineer who will have to be
equipped with a wetsuit and fins to do maintenance.
Actually looks like Microsoft have very sensibly filled the thing with
a dried nitrogen gas, which makes a lot of sense. And it is supposed
to be maintenance free, I would imagine any degraded servers will just
be switched off.
Prof Ian Bitterlin says "You just end up with a warmer sea and bigger
fish,"
I have told the tale on here before about the town I grew up in which
had a huge Singer factory. The factory had its own power station which
discharged hot water into the local canal. The canal was famous for
having foot long goldfish.
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

[Douglas Eadline]

-snip-

i'm not sure i see a point in all this anyhow, it's a neat science
experiment, but what's the ROI on sinking a container full of servers
vs just pumping cold seawater from 100ft down

I had the same thought. You could even do a salt water/clear water
heat exchange and not have the salt water near the servers.

From a risk perspective, failure under 100 ft of sea water
would seem to much more catastrophic vs failure on land and
cooling with pumped water (maybe I read too much N.N. Taleb).

[Joe Landman]

Post by Douglas Eadline
-snip-

i'm not sure i see a point in all this anyhow, it's a neat science
experiment, but what's the ROI on sinking a container full of servers
vs just pumping cold seawater from 100ft down

I had the same thought. You could even do a salt water/clear water
heat exchange and not have the salt water near the servers.
From a risk perspective, failure under 100 ft of sea water
would seem to much more catastrophic vs failure on land and
cooling with pumped water (maybe I read too much N.N. Taleb).

Imagine 100kW or so ... suddenly discovering that the neat little hole
in the pipe enables this highly conductive ionic fluid to short ...
somewhere between 1V and 12V DC.  10's to 100's of thousands of Amps.  I
wouldn't wanna be anywhere near that when it lets go.
--
Joe Landman
e: ***@gmail.com
t: @hpcjoe
w: https://scalability.org
g: https://github.com/joelandman
l: https://www.linkedin.com/in/joelandman

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http

[Chris Samuel]

The report interestingly makes a comparison to cruise lines and the US Navy
having large IT infrastructures at sea.

Some oil & gas companies have HPC systems onboard their survey vessels to
process data at sea. One example from Australia:

https://www.nextplatform.com/2018/01/08/hpc-optimizes-energy-exploration-oil-gas-startups/

# DUG provides hardware and software to marine geophysical company Polarcus*,
# fitting out its fleet of seagoing vessels for marine seismic data acquisition.
# Dr. Stuart Midgley, DUG’s systems architect, notes that space is very
# limited on each vessel, so the onboard supercomputing system must be
# extremely powerful for data processing and imaging, yet consume minimal
# electrical power and occupy a limited footprint.

cheers.
Chris

[Prentice Bisbal]

Post by Chris Samuel

The report interestingly makes a comparison to cruise lines and the US Navy
having large IT infrastructures at sea.

Some oil & gas companies have HPC systems onboard their survey vessels to
https://www.nextplatform.com/2018/01/08/hpc-optimizes-energy-exploration-oil-gas-startups/
# Dr. Stuart Midgley, DUG’s systems architect,

Wait a minute... I think I've heard that name before...
_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/

[Jonathan Aquilina]

If I’m not mistaken said person mention is a subscriber on this list

Sent from my iPhone

Post by Prentice Bisbal

Post by Chris Samuel

The report interestingly makes a comparison to cruise lines and the US Navy
having large IT infrastructures at sea.

Some oil & gas companies have HPC systems onboard their survey vessels to
https://www.nextplatform.com/2018/01/08/hpc-optimizes-energy-exploration-oil-gas-startups/
# Dr. Stuart Midgley, DUG’s systems architect,

Wait a minute... I think I've heard that name before...
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.b

[John Hearns via Beowulf]

Stuart Midgley works for DUG? They are currently recruiting for an HPC
manager in London... Interesting...

Post by Chris Samuel

The report interestingly makes a comparison to cruise lines and the US

Navy

having large IT infrastructures at sea.

Some oil & gas companies have HPC systems onboard their survey vessels to
https://www.nextplatform.com/2018/01/08/hpc-optimizes-
energy-exploration-oil-gas-startups/
# DUG provides hardware and software to marine geophysical company Polarcus*,
# fitting out its fleet of seagoing vessels for marine seismic data acquisition.
# Dr. Stuart Midgley, DUG’s systems architect, notes that space is very
# limited on each vessel, so the onboard supercomputing system must be
# extremely powerful for data processing and imaging, yet consume minimal
# electrical power and occupy a limited footprint.
cheers.
Chris
--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

[Christopher Samuel]

Post by John Hearns via Beowulf
Stuart Midgley works for DUG?

Yup, for over a decade.. :-)

[Lux, Jim (337K)]

Post by Douglas Eadline
-snip-

i'm not sure i see a point in all this anyhow, it's a neat science
experiment, but what's the ROI on sinking a container full of servers
vs just pumping cold seawater from 100ft down

I had the same thought. You could even do a salt water/clear water
heat exchange and not have the salt water near the servers.
From a risk perspective, failure under 100 ft of sea water
would seem to much more catastrophic vs failure on land and
cooling with pumped water (maybe I read too much N.N. Taleb).

Imagine 100kW or so ... suddenly discovering that the neat little hole
in the pipe enables this highly conductive ionic fluid to short ...
somewhere between 1V and 12V DC. 10's to 100's of thousands of Amps. I
wouldn't wanna be anywhere near that when it lets go.



Ahem, that is NOT the use case this kind of capability is aiming at. You're right - pumps and pipes and hoses are cheaper. If you're close to the coastline. What if you want your "datacenter" somewhere on a line between Iceland and the Hebrides. (sorry, I just re-read Hunt for Red October on a long plane flight)

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://

[Lux, Jim (337K)]

That is a most excellent book. And it brings to mind some of the more complex aspects of maintenance for that cluster.

I think that’s actually an important area for cluster development – most of the work, to date, has been in building high performance computing in environments that are easily accessible. *I* have always been interested in HPC in inaccessible or rugged environments. If you want a cluster computer at Europa, you need reliability and remote maintainability, without a full duplex ethernet connection with low latency. If you want to do hard core signal processing (say archaeological radar or seismic processing) in the jungle, where your connection to home is a 9600 bps Iridium satphone.

While there are interesting engineering challenges in building a 1000 core cluster in a building to which you can back a truck up to; now think about how you’d build/manage/repair that cluster when the light time delay is 30 minutes, and it takes a year to get there.



From: Beowulf <beowulf-***@beowulf.org> on behalf of "***@beowulf.org" <***@beowulf.org>
Reply-To: John Hearns <***@googlemail.com>
Date: Thursday, June 7, 2018 at 8:39 AM
To: "***@beowulf.org" <***@beowulf.org>
Subject: Re: [Beowulf] Fwd: Project Natick

The report interestingly makes a comparison to cruise lines and the US Navy having large IT infrastructures at sea.
I guess cruise ships of course have servers plus satcomms, as do warships.
But the thought of the SOSUS sonar chain comes to mind... then again those electronics will be down a lot deeper than this.
Though I am sure a few racks of FPGAs near your SOSUS listening devices would be good...

Going wildly off topic as usual this book https://www.nytimes.com/1998/11/08/us/a-tale-of-daring-american-submarine-espionage.html
about Operation Ivy Bells is fantastic https://en.wikipedia.org/wiki/Operation_Ivy_Bells

[Chris Samuel]

Post by Lux, Jim (337K)
If you want a cluster computer at Europa, you need reliability and remote
maintainability

I suspect you could probably find some volunteers for on-site work... ;-)

[Prentice Bisbal]

Post by Chris Samuel

Post by Lux, Jim (337K)
If you want a cluster computer at Europa, you need reliability and remote
maintainability

I suspect you could probably find some volunteers for on-site work... ;-)

I know some people I'd like to volunteer for that. ;-)

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beow

[Lux, Jim (337K)]

Oh, the radiation dose rate on the surface (variously given as 1-100 Rad/second (0.01-1 Gy/s), or 5.4 Sv/day (which are orders of magnitude different) means the first "jupiter rise" would be spectactular, and then you'd die. 5 Sv (500 rem) is pretty much a lethal dose.

Post by Lux, Jim (337K)
If you want a cluster computer at Europa, you need reliability and remote
maintainability

I suspect you could probably find some volunteers for on-site work... ;-)

--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf


_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) vi

[Chris Samuel]

Post by Lux, Jim (337K)
Oh, the radiation dose rate on the surface (variously given as 1-100
Rad/second (0.01-1 Gy/s), or 5.4 Sv/day (which are orders of magnitude
different) means the first "jupiter rise" would be spectactular, and then
you'd die. 5 Sv (500 rem) is pretty much a lethal dose.

Ouch. OK, I'll pass on that...

[Lux, Jim (337K)]

Well, it's all about shielding mass.

If you melt yourself down a meter into the ice, then you're pretty well shielded.

So that big steel tank for Project Natick might work ok.



Jim Lux
[phone] (office)
[phone] (cell)


-----Original Message-----
From: Beowulf [mailto:beowulf-***@beowulf.org] On Behalf Of Chris Samuel
Sent: Tuesday, June 12, 2018 5:34 AM
To: ***@beowulf.org
Subject: Re: [Beowulf] Fwd: Project Natick

Post by Lux, Jim (337K)
Oh, the radiation dose rate on the surface (variously given as 1-100
Rad/second (0.01-1 Gy/s), or 5.4 Sv/day (which are orders of magnitude
different) means the first "jupiter rise" would be spectactular, and
then you'd die. 5 Sv (500 rem) is pretty much a lethal dose.

Ouch. OK, I'll pass on that...

[Chris Samuel]

I imagine it would have to be filtered, too, to keep small marine life
and debris from clogging up the piping. I wonder if any forms of marine
life in that part of the ocean would like the warm water inside the
heat exchangers or at the exhaust and try to make it their homes.

Seaweed and jellyfish have been known to cause issues for Scottish nuclear
reactors, so I suspect this wouldn't be immune to those worries either!

http://www.heraldscotland.com/news/16066525.Scots_nuclear_reactor_shut_down_over_seaweed_concern/

All the best,
Chris

[Lux, Jim (337K)]

Where, exactly, is this... I spent a week in Orkney a few years ago (we drove up from Glasgow, also an interesting proposition) - it's worth a visit for the prehistoric archaeology (largest prehistoric settlement still preserved at Skara Brae, for instance, but realistically the dig at the Ness of Brodgar (http://www.nessofbrodgar.com/) is more interesting. And, of course in the summer, it never really gets dark at night. Don't go in the Winter.

Did they sink it off the northern coast in Kirkwall or on the south side at Scapa Flow? (where there's plenty of sunken ships to keep it company)

They talk about how harsh the environment is - yeah, there's not many trees on the islands, but the water is surprisingly warm, and the waves aren't all that big. It's nothing like being on an oil rig in the North Sea, for instance (I've not been there, but have heard stories). Scapa Flow is apparently a big scuba diving destination, for instance.

Sounds a bit like a "let's make a good PR story out of this" - and with convenient access to civilization - Orkney is a whole lot more comfortable than a lot of other places off the coast of Scotland (or France, for that matter)

I imagine it would have to be filtered, too, to keep small marine life
and debris from clogging up the piping. I wonder if any forms of marine
life in that part of the ocean would like the warm water inside the
heat exchangers or at the exhaust and try to make it their homes.

Seaweed and jellyfish have been known to cause issues for Scottish nuclear
reactors, so I suspect this wouldn't be immune to those worries either!

http://www.heraldscotland.com/news/16066525.Scots_nuclear_reactor_shut_down_over_seaweed_concern/

All the best,
Chris
--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf


_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinf

[Chris Samuel]

Post by Lux, Jim (337K)
Where, exactly, is this... I spent a week in Orkney a few years ago (we
drove up from Glasgow, also an interesting proposition) - it's worth a
visit for the prehistoric archaeology (largest prehistoric settlement still
preserved at Skara Brae, for instance, but realistically the dig at the
Ness of Brodgar (http://www.nessofbrodgar.com/) is more interesting.

+lots (it's on my bucket list)

Off topic - if you are interested in UK archaeology then I can strongly
recommend Current Archaeology magazine (they have been covering the various
digs in Orkney over the past few years). Their website has this review of the
state of Orkney archaeology from 2016 (which includes the discovery that
timber buildings predate the familiar stone buildings there).

https://www.archaeology.co.uk/articles/features/a-tale-of-two-neolithics.htm

Post by Lux, Jim (337K)
Did they sink it off the northern coast in Kirkwall or on the south side at
Scapa Flow? (where there's plenty of sunken ships to keep it company)

Neither, it's at the Billia Croo wave test site which is on the western side
of Mainland:

http://www.emec.org.uk/facilities/wave-test-site/

# Subjected to the powerful forces of the North Atlantic Ocean, it is an area
# with one of the highest wave energy potentials in Europe with an average
# significant wave height of 2 – 3 metres, but reaching extremes of up to 17m
# (the highest wave recorded by EMEC so far). The site consists of five
# cabled test berths in up to 70m water depth (four at 50m, one deeper),
# located approximately 2km offshore and 0.5km apart. In addition to this, a
# near shore berth is situated closer to the substation for shallow water
# projects.

PDF map here:

http://www.emec.org.uk/?wpfb_dl=164

cheers!
Chris

[Jonathan Engwall]

The whitepaper says the acceleration processes 70 images per second per
watt.
More juice more better?
I will read the whole thing a little later tonight.

Reading into this a bit more on the Microsoft site, the intention is to
power these things using renewables such as wind or tidal power.
I've never been to Orkney, but as it famously has no trees ther eis

plenty

of wind I am sure...
Might make sense actually as they say for remote communities.
The cynic in me says why in the heck would a remote community NEED 12 or
more racks of servers, but this wont be for local use.
It makes sense in terms of having free cost power (OK - I know the true
cost is the construction of a wind turbine or two) and free cost cooling.
The total power is 240kW which is a respectable amount of power - not as
dense as big HPC installations these days, but pretty respectable.

I maybe will read the source article if I find time, but from words of
computer + salt water + wind turbine + Microsoft
I suggest it will not go anywhere and is in fact some kind of
marketing stunt.
But I will wait a year and see if there is a reason to change my
opinion.

One sincerely hopes that if things like this do get deployed in the ocean
then the steel module, the wind turbine and the servers are recycled at

the

end of life and not just abandoned.

Yeah.
--
Regards,
Tomasz Rola
--
** A C programmer asked whether computer had Buddha's nature. **
** As the answer, master did "rm -rif" on the programmer's home **
** directory. And then the C programmer became enlightened... **
** **
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

[Jonathan Engwall]

The whitepaper describes an association between elements of an image with
elements that surround that element.
The black stripes on a tiger for example distinguish from all the other
orange animals. If there are any.
They call this a CNN.
The fpga stores data on a chip, but the servers have teslas. So I don't
have a guess what an fpga might look like.

Post by Prentice Bisbal

Post by Prentice Bisbal
Has anybody seen any more details on how the cooling actually occurs

withing

Post by Prentice Bisbal
the capsule?

https://datacenterfrontier.com/the-watery-edge-microsoft-
deploys-undersea-servers-in-scotland/
# A key change from the prototype was in the cooling system, where Naval
Group
# adapted a heat-exchange process commonly used for cooling submarines,
# piping seawater directly through the radiators on the back of each of
the 12
# server racks and back out into the ocean.
So water cooled doors, but presumably hardened against the corrosive
properties of sea water?
Going after the bitcoin crowd perhaps?
cheers,
Chris
--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

[John Hearns via Beowulf]

Thinking about submarines, I mentioned a UK secure site on another thread.
That site may or may not have been something to do with submarines.
I have never been on board a submarine, however if I was faced with the
problem of cooling on board one I would shy away from what that article
implies,
ie opening holes in the pressure hull and pumping water around, even if you
are heat exchanging with sea water and distilled water in the actual racks.
I would think about something using the pressure hull as the radiator and
using natural convection.
Then again, those pcitures show a smooth metal cylinder, with nothing
looking like fins or radiators on the side.
But to me introducing valves into the side of a structure underwater is
just asking for failure.
Maybe the seawater enters a heat exchanger interior which is completely
sealed off from the rest of the structure.

Post by Jonathan Engwall
The whitepaper describes an association between elements of an image with
elements that surround that element.
The black stripes on a tiger for example distinguish from all the other
orange animals. If there are any.
They call this a CNN.
The fpga stores data on a chip, but the servers have teslas. So I don't
have a guess what an fpga might look like.

Post by Prentice Bisbal

Post by Prentice Bisbal
Has anybody seen any more details on how the cooling actually occurs

withing

Post by Prentice Bisbal
the capsule?

https://datacenterfrontier.com/the-watery-edge-microsoft-dep
loys-undersea-servers-in-scotland/
# A key change from the prototype was in the cooling system, where Naval
Group
# adapted a heat-exchange process commonly used for cooling submarines,
# piping seawater directly through the radiators on the back of each of
the 12
# server racks and back out into the ocean.
So water cooled doors, but presumably hardened against the corrosive
properties of sea water?
Going after the bitcoin crowd perhaps?
cheers,
Chris
--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

[Lawrence Stewart]

Has anybody seen any more details on how the cooling actually occurs withing
the capsule?

https://datacenterfrontier.com/the-watery-edge-microsoft-deploys-undersea-servers-in-scotland/
# A key change from the prototype was in the cooling system, where Naval Group
# adapted a heat-exchange process commonly used for cooling submarines,
# piping seawater directly through the radiators on the back of each of the 12
# server racks and back out into the ocean.
So water cooled doors, but presumably hardened against the corrosive
properties of sea water?
Going after the bitcoin crowd perhaps?

Microsoft has been interested in FPGAs for datacenters for years. The work is quite interesting and a review link is https://queue.acm.org/detail.cfm?id=3231573 <https://queue.acm.org/detail.cfm?id=3231573>
They use it for search acceleration, among other things.
Other purposes i’ve heard about include a place to keep customer crypto keys inaccessible to the service operator, and yes, making the hardware available to customers.

cheers,
Chris
--
Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

[Chris Samuel]

And I'm definitely not an expert on Orkney, but I did work with a guy
from Scotland, and I'm pretty sure he had stories about how sparsely
populated Orkney was due to the rugged terrain and inhospitable weather,
so this test case isn't exactly near a large population center, either.

I suspect the main driver for having the prototypes in Scotland is the fact
that it's located at an existing marine testing facility for renewable energy
development, so plenty of expertise on hand plus lower latency to the UK and
Europe compared to comparable sites in the US.

[Prentice Bisbal]

Post by Chris Samuel

And I'm definitely not an expert on Orkney, but I did work with a guy
from Scotland, and I'm pretty sure he had stories about how sparsely
populated Orkney was due to the rugged terrain and inhospitable weather,
so this test case isn't exactly near a large population center, either.

I suspect the main driver for having the prototypes in Scotland is the fact
that it's located at an existing marine testing facility for renewable energy
development, so plenty of expertise on hand plus lower latency to the UK and
Europe compared to comparable sites in the US.

I knew that. I was just being sarcastic and poking fun at this whole
stunt. ;)
_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://ww

[Prentice Bisbal]

I should also add that Chris' URL provides a much better shot of the
racks going into the capsule, where you can see heat exchangers and
their fans are incorporate into the rack structure itself. In the
picture I commented on yesterday, you couldn't see that. It just looked
like the computer racks from that angle.

Has anybody seen any more details on how the cooling actually occurs withing
the capsule?

https://datacenterfrontier.com/the-watery-edge-microsoft-deploys-undersea-servers-in-scotland/
# A key change from the prototype was in the cooling system, where Naval Group
# adapted a heat-exchange process commonly used for cooling submarines,
# piping seawater directly through the radiators on the back of each of the 12
# server racks and back out into the ocean.
So water cooled doors, but presumably hardened against the corrosive
properties of sea water?
Going after the bitcoin crowd perhaps?
cheers,
Chris

_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http:/

[John Hearns via Beowulf]

Reading into this a bit more on the Microsoft site, the intention is to
power these things using renewables such as wind or tidal power.
I've never been to Orkney, but as it famously has no trees ther eis plenty
of wind I am sure...

Might make sense actually as they say for remote communities.
The cynic in me says why in the heck would a remote community NEED 12 or
more racks of servers, but this wont be for local use.
It makes sense in terms of having free cost power (OK - I know the true
cost is the construction of a wind turbine or two) and free cost cooling.
The total power is 240kW which is a respectable amount of power - not as
dense as big HPC installations these days, but pretty respectable.

One sincerely hopes that if things like this do get deployed in the ocean
then the steel module, the wind turbine and the servers are recycled at the
end of life and not just abandoned.

Post by Prentice Bisbal
I heard about this on BBC World News this morning on my way into work. I
waas going to share this here myself this morning.
What isn't clear is how the heat is being transferred from the CPUs to the
seawater. My best guess at the moment is that the capsule's steel walls
conduct heat from the hot air (nitrogen gas, actually), to the seawater,
cooling the ambient "air", and then the air is circulated just using the
chassis fans. It's possible there are some supplemental fans to circulate
air around the capsule, but based on the photos of the racks being loaded
into the capsule, it doesn't look like there'd be much room for that. This
would mean the processors are still air-cooled themselves.
Has anybody seen any more details on how the cooling actually occurs
withing the capsule?
Payload 12 racks containing 864 standard Microsoft datacenter servers
with FPGA acceleration and 27.6 petabytes of disk. This Natick datacenter
is as powerful as several thousand high end consumer PCs and has enough
storage for about 5 million movies.
Since they are equipped with FPGAs, it doesn't sound like they're doing
routine workloads, so this technique might not be transferable to you or
me. I'm assuming the FPGAs will get much better performance per watt than a
general processor, reducing the heatload in the capsule vs. doing the same
workload with only x86 processors. Does any one know what the intended
workload of this system is?
https://www.bbc.com/news/technology-44368813
https://natick.research.microsoft.com/
I must admit my first thoughts on hearing an item about this on Radio
Scotland is that now that humans have laid waste to the surface of the
Earth we are going to boil the oceans.
My second thought is for the poor HPC engineer who will have to be
equipped with a wetsuit and fins to do maintenance.
Actually looks like Microsoft have very sensibly filled the thing with a
dried nitrogen gas, which makes a lot of sense. And it is supposed to be
maintenance free, I would imagine any degraded servers will just be
switched off.
Prof Ian Bitterlin says "You just end up with a warmer sea and bigger
fish,"
I have told the tale on here before about the town I grew up in which had
a huge Singer factory. The factory had its own power station which
discharged hot water into the local canal. The canal was famous for having
foot long goldfish.
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
_______________________________________________
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf

[Lux, Jim (337K)]

I imagine it would have to be filtered, too, to keep small marine life
and debris from clogging up the piping.

Filtered through a Millipore filter? Keeping things from growing on your underwater stuff is tough. Of course, you could just make it out of copper, good thermal conductivity, hostile to marine growth.


I wonder if any forms of marine
life in that part of the ocean would like the warm water inside the
heat exchangers or at the exhaust and try to make it their homes.

Post by Prentice Bisbal
Going after the bitcoin crowd perhaps?

Or, just for marketing (speaking as a cynical kind of dude, sometimes) - "Our nodes have FPGA, not like those inferior nodes from competitor X"

I can't imagine a bitcoin miner "buying compute cycles" - Do it by making a "deal" with the government to give you cheap joules and share the proceeds either above or below the table.



_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/l

[Lux, Jim (337K)]

I imagine it would have to be filtered, too, to keep small marine life and
debris from clogging up the piping. I wonder if any forms of marine life in
that part of the ocean would like the warm water inside the heat exchangers
or at the exhaust and try to make it their homes.

my guess it's probably a low risk. not only are the pipes likely full
copper, which is toxic to most marine life, but the flow rate inside
the pipes is probably high enough that nothing has much of a chance to
stick. there's probably just some course basic filters that need to
scrubbed clear every once in a while.

i'm not sure i see a point in all this anyhow, it's a neat science
experiment, but what's the ROI on sinking a container full of servers
vs just pumping cold seawater from 100ft down
---
Yes, copper is your friend.

I like the idea (from Michigan?) - have a huge pit full of pipes next to the server center and spray water during the winter to form ice, then melt the ice during the summer.
Pumping cold water around is *much* easier than sinking a server farm in the ocean.

On the other hand, there *are* people who would be interested in a sea bottom computational capability to process data from, oh, an array of pressure, acoustic, and other sensors, so that the link to the surface doesn't have to carry a huge volume of data. Tsunami detection, for instance, or tracking sea life migration.



_______________________________________________
Beowulf mailing list, ***@beowulf.org sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/l