Quote from: Selerius on March 15, 2017, 01:50:22 AM

Quote from: clivemy on March 14, 2017, 08:41:43 PM

Quote from: Selerius on March 14, 2017, 02:21:49 PM

Quote from: Bgjjj2016 on March 14, 2017, 06:43:34 AM

Quote from: Evil-Knievel on March 14, 2017, 06:10:22 AM

[...]

Hi EK, always good to hear from you. Thanks for the updates.
Maybe I misunderstood, please forgive me in advance. Do we need any funding for any equipment you need to test the SN ?

Hardware requirements for a SN are not that high. I don't speak for EK, but there is practical certainty he has it totally covered.

I've been wondering what the specs would be to run an SN, now core software & SN.

Can I ask what you think would be the specs for a system to run the software, approximately.

As far as I know, supernodes are supposed to be reliably online, maintain a large XEL deposit as a sign of relative trustworthiness, and accept running novel foreign code (an aspect which should be safe in the restricted language used but still such that some places would feel more comfortable just running a regular node instead).

Yet the actual hardware requirements are apparently not really super extreme. At least back on January 24th, EK made a reference to work using 5 gigabytes of RAM memory (on top of ordinary use by the operating system and so on, of course).

Since 16GB and 32GB RAM desktops are not uncommon among recent models today, that wouldn't be much at all. The supernode vs regular node hardware requirements difference may not be "super extreme" versus "ordinary" so much as "mildly high" versus "okay even if really mediocre and outdated." Of course, the adjectives are my own, not EK's, not a quote.

I don't know if exactly the 5 GB example is still representative.

But my earlier comment was mainly just based on the following: I think EK is coding supernodes to run on what he already has now, as the only easy way to do development and testing. He is not programming them in a manner which would require him to get much different hardware. There would be no reason to do the latter, as he chooses the expected complexity and memory requirements of work packages, which were originally going to be scaled down to even relatively mediocre computers.

For a fully specific answer with exact numbers, we might need to wait for EK to announce it when he is done coding. I am no real expert, just a bystander. Yet I bet it will be that just about any 16GB RAM (or maybe even 8GB) computer with permanent internet access could do it, if willing to have a large portion of system resources consumed running the software.

I'd say, the requirements for all SNs would need to be high enough to make it reliably process all network jobs. Obviously, it highly depends on the number of SNs, number of regular nodes, and the network workload. I guess EK has an idea how to balance all of this, otherwise he probably wouldn't implement this idea right?