Author Topic: Question with networking computers  (Read 425 times)

0 Members and 1 Guest are viewing this topic.

Online Desert Fox

  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 13885
  • Hopeful Non-Theist
    • Kitsune's Web Page
Question with networking computers
« on: April 19, 2017, 05:19:43 PM »
Let us assume that you want to network a group of computers for more processing power. The catch is that all of these computers are in spaceships that are light seconds apart so no near instantaneous communications. Could they be networked in a fashion for greater processing power in such a case or would the communication issue create problems?
"Give me the storm and tempest of thought and action, rather than the dead calm of ignorance and faith. Banish me from Eden when you will; but first let me eat of the fruit of the tree of knowledge."
— Robert G. Ingersoll

Offline The Latinist

  • Cyber Greasemonkey
  • Technical Administrator
  • Reef Tank Owner
  • *****
  • Posts: 9264
Re: Question with networking computers
« Reply #1 on: April 19, 2017, 05:36:13 PM »
That depends on what you're trying to do: if it's data processing that could be broken down into chunks you could do it the way a distributed computing project does.  The problem would be if there were a lot of I/O or threading that would require locks, etc. If I have to wait for a signal to travel between ships before starting on the next step of the problem, that could easily outweigh any gains from parallelism.
I would like to propose...that...it is undesirable to believe in a proposition when there is no ground whatever for supposing it true. — Bertrand Russell

Online 2397

  • Seasoned Contributor
  • ****
  • Posts: 831
Re: Question with networking computers
« Reply #2 on: April 19, 2017, 05:37:24 PM »
Like Folding@home, SETI@home, Bitcoin, BitTorrent (sort of)? Nothing is quite lightseconds apart on Earth, but if you can split up the work in a way that there's no overlap, and if it's okay that the work is completed at different times, then I don't see the issue.

I'd be surprised that this was necessary for a series of spaceships to do, but I guess there's no limit to the variety of events that "spaceships" could be dealing with.

For a different analogy, we have instruments all over the Solar System that people work on all day (and sometimes night, e.g. going by Mars time), and they have to deal with up to hours of delay.

Edit: Overlapping work would be okay if you had a time sensitive problem to solve that it would take a different amount of time each time to solve. Or if you can have each computer try a different way of solving a problem, when you don't know which way is the fastest. Meaning that one of the computers might solve the problem faster than the others, but you don't know which one until it gives you the answer.
« Last Edit: April 19, 2017, 06:00:35 PM by 2397 »

Online Desert Fox

  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 13885
  • Hopeful Non-Theist
    • Kitsune's Web Page
Re: Question with networking computers
« Reply #3 on: April 21, 2017, 04:27:22 PM »
It happened in a sci-fi book I was reading where they used the entire computer network of a fleet to model the effects of a type of explosion to try to prevent it. . . . .Lost Fleet series
"Give me the storm and tempest of thought and action, rather than the dead calm of ignorance and faith. Banish me from Eden when you will; but first let me eat of the fruit of the tree of knowledge."
— Robert G. Ingersoll

Offline The Latinist

  • Cyber Greasemonkey
  • Technical Administrator
  • Reef Tank Owner
  • *****
  • Posts: 9264
Re: Question with networking computers
« Reply #4 on: April 21, 2017, 06:24:45 PM »
That sounds like the kind of problem that might be hard to break up into smaller parts.
I would like to propose...that...it is undesirable to believe in a proposition when there is no ground whatever for supposing it true. — Bertrand Russell

Online Desert Fox

  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 13885
  • Hopeful Non-Theist
    • Kitsune's Web Page
Re: Question with networking computers
« Reply #5 on: April 21, 2017, 06:54:52 PM »
What do you think the largest lag one could deal with? Maybe you could pull the heavies within a fraction of a light second apart to network them? A light second after all is about 300,000 km.
"Give me the storm and tempest of thought and action, rather than the dead calm of ignorance and faith. Banish me from Eden when you will; but first let me eat of the fruit of the tree of knowledge."
— Robert G. Ingersoll

Offline estockly

  • Stopped Going Outside
  • *******
  • Posts: 5013
Re: Question with networking computers
« Reply #6 on: April 21, 2017, 11:19:40 PM »
Like Folding@home, SETI@home, Bitcoin, BitTorrent (sort of)? Nothing is quite lightseconds apart on Earth...

There certainly is, if we're talking about transmission over the internet.  The physical distance between the systems is irrelevant, even the physical distance the signal has to travel (which could conceivably be light seconds) is irrelevant. What's relevant is the transmission time. All of those distributive computing services you listed (and probably all the others) are designed to deal with significant lags in transmission throughput. They were designed to send and receive complete data packets. Complete raw packets go out, complete, processed packets are returned.

So, the short answer is: Yes, the DFFS (Desert Fox's Fleet of Spaceships) could easily be networked for distributed computing, there should be no issues with transmission (assuming adequate bandwidth).
and Donald Trump is President of the United States.


"Well, sometimes the magic works. Sometimes it doesn't." -- Chief Dan George, "Little Big Man"

Offline Caffiene

  • Stopped Going Outside
  • *******
  • Posts: 4768
Re: Question with networking computers
« Reply #7 on: April 22, 2017, 02:03:10 AM »
That sounds like the kind of problem that might be hard to break up into smaller parts.

Depends how its being modeled. If its approximated in a fashion similar to weather forecasting where you break it up into cells it would probably be possible. Or if you could break it down into sub-equations like Fourier Transforms for sound and solve individually. In fiction it could be explained away pretty easily.
[Lurk Mode Disengage]

Online 2397

  • Seasoned Contributor
  • ****
  • Posts: 831
Re: Question with networking computers
« Reply #8 on: April 22, 2017, 04:18:52 AM »
There certainly is, if we're talking about transmission over the internet.  The physical distance between the systems is irrelevant, even the physical distance the signal has to travel (which could conceivably be light seconds) is irrelevant. What's relevant is the transmission time. All of those distributive computing services you listed (and probably all the others) are designed to deal with significant lags in transmission throughput. They were designed to send and receive complete data packets. Complete raw packets go out, complete, processed packets are returned.

Right, but lighseconds describes a physical distance. Alternatively you could take into account the given medium, to be pedantic. So unless there's a wiring setup that makes it so that a signal has to travel the equivalent of 299 792 458 meters in a vacuum, double that for plural, then it was an accurate statement.

The biggest segments of the global network are the undersea cables, which leaves a lot less potential for multiplying the distance. Of course they're not completely straight, or level, so that adds some distance. But I have a difficulty imagining how any networked computers would end up with effective distances nearly 15 times the longest direct distance between two points on Earth.
« Last Edit: April 22, 2017, 04:21:54 AM by 2397 »

Online Noisy Rhysling

  • Well Established
  • *****
  • Posts: 1109
    • Hyperwar, WWII in Hypertext.
Re: Question with networking computers
« Reply #9 on: April 22, 2017, 08:19:44 AM »
"Packet passing". Each ship's computer is "highly qualified" in one or more routines. When the ship gets a packet the computer looks to see if it can help with the problem. If not it looks for the next nearest ship that is HQ for that problem or something close. "Gatekeeper" computers at intervals to make sure the HQ computers haven't gone off the rails, and puts things back on track if they have. "Master" computers check up on the Gatekeepers and assemble the packets. They "assemble" the solved portions of the problem and report unsolved portions to the Master computers. The Master computers decide if the groups should come together to instantly network and create the Mega computer to provide maximum available computing power.

Etc.
"Sunday's horoscope is note worthy because of its strange, sudden and wholly unpredictable and inexplicable occurrences, affecting all phases of life." Your Horoscope" L.A. Evening Herald Express, Sat, 12/06/41

Offline The Latinist

  • Cyber Greasemonkey
  • Technical Administrator
  • Reef Tank Owner
  • *****
  • Posts: 9264
Re: Question with networking computers
« Reply #10 on: April 22, 2017, 09:02:05 AM »
Ultimately, as I said from the beginning, it all comes down to whether the problem can be appropriately modeled for a distributed computing system with a significant lag in communications between threads. 
I would like to propose...that...it is undesirable to believe in a proposition when there is no ground whatever for supposing it true. — Bertrand Russell

Offline estockly

  • Stopped Going Outside
  • *******
  • Posts: 5013
Re: Question with networking computers
« Reply #11 on: April 22, 2017, 11:59:30 AM »
There certainly is, if we're talking about transmission over the internet.  The physical distance between the systems is irrelevant, even the physical distance the signal has to travel (which could conceivably be light seconds) is irrelevant. What's relevant is the transmission time. All of those distributive computing services you listed (and probably all the others) are designed to deal with significant lags in transmission throughput. They were designed to send and receive complete data packets. Complete raw packets go out, complete, processed packets are returned.

Right, but lighseconds describes a physical distance. Alternatively you could take into account the given medium, to be pedantic. So unless there's a wiring setup that makes it so that a signal has to travel the equivalent of 299 792 458 meters in a vacuum, double that for plural, then it was an accurate statement.

The biggest segments of the global network are the undersea cables, which leaves a lot less potential for multiplying the distance. Of course they're not completely straight, or level, so that adds some distance. But I have a difficulty imagining how any networked computers would end up with effective distances nearly 15 times the longest direct distance between two points on Earth.

Again, the actual distance is irrelevant. If the DFFS are light seconds apart, the only challenge that creates is the delay between a packet being sent and received. That kind of delay occurs all the time on the terrestrial internet now, and is accounted for in distributed computing systems. I would say distributed computing over the internet wouldn't work at all if delays of seconds between sending and receiving weren't accounted for.


and Donald Trump is President of the United States.


"Well, sometimes the magic works. Sometimes it doesn't." -- Chief Dan George, "Little Big Man"

Offline estockly

  • Stopped Going Outside
  • *******
  • Posts: 5013
Re: Question with networking computers
« Reply #12 on: April 22, 2017, 02:28:15 PM »
"Packet passing". Each ship's computer is "highly qualified" in one or more routines.


More likely each ship has multiple computers, and when any are not being used for their primary task they are given packets to process. So each ship would have its own network and dozens/hundreds/thousands of processors churning away




Your mileage may vary.
and Donald Trump is President of the United States.


"Well, sometimes the magic works. Sometimes it doesn't." -- Chief Dan George, "Little Big Man"

Online Noisy Rhysling

  • Well Established
  • *****
  • Posts: 1109
    • Hyperwar, WWII in Hypertext.
Re: Question with networking computers
« Reply #13 on: April 22, 2017, 06:19:42 PM »
"Packet passing". Each ship's computer is "highly qualified" in one or more routines.


More likely each ship has multiple computers, and when any are not being used for their primary task they are given packets to process. So each ship would have its own network and dozens/hundreds/thousands of processors churning away




Your mileage may vary.
Did that on the fly. However, like a modern day battlefleet the ships could be of various size, allowing for more or less space for computers. Save ten "destroyers" computer power is equal to one "cruiser", and ten cruisers equal to... etc.
"Sunday's horoscope is note worthy because of its strange, sudden and wholly unpredictable and inexplicable occurrences, affecting all phases of life." Your Horoscope" L.A. Evening Herald Express, Sat, 12/06/41

Offline estockly

  • Stopped Going Outside
  • *******
  • Posts: 5013
Re: Question with networking computers
« Reply #14 on: April 22, 2017, 07:26:54 PM »
"Packet passing". Each ship's computer is "highly qualified" in one or more routines.


More likely each ship has multiple computers, and when any are not being used for their primary task they are given packets to process. So each ship would have its own network and dozens/hundreds/thousands of processors churning away




Your mileage may vary.
Did that on the fly. However, like a modern day battlefleet the ships could be of various size, allowing for more or less space for computers. Save ten "destroyers" computer power is equal to one "cruiser", and ten cruisers equal to... etc.

Sure, but I figure if they're advanced enough to have a fleet of spacecraft large enough to benefit from distributed computing they're probably advanced enough to make processors very small and powerful.
and Donald Trump is President of the United States.


"Well, sometimes the magic works. Sometimes it doesn't." -- Chief Dan George, "Little Big Man"

 

personate-rain