Working on multisystem clustered mining
Codename: Satan's Workshop
trying to build 1.0.6 on a new image Im creating and Im getting this
INFO: Analysed target //:PoolMiner (55 packages loaded). INFO: Found 1 target... ERROR: missing input file '@local_jdk//:jre/lib/rt.jar' ERROR: /home/ubuntu/snowblossom-1.0.6/miner/BUILD:3:1: //miner:miner: missing input file '@local_jdk//:jre/lib/rt.jar' Target //:PoolMiner failed to build Use --verbose_failures to see the command lines of failed build steps. ERROR: /home/ubuntu/snowblossom-1.0.6/miner/BUILD:3:1 1 input file(s) do not exist INFO: Elapsed time: 10.601s, Critical Path: 3.19s INFO: 11 processes: 11 linux-sandbox. FAILED: Build did NOT complete successfully
figured it out, missed a step install bazel
ha
Which pool is this?
it has one miner, apparently
so I imagine some private operation
I'm going to say, lobsters from deep in the atlantic
heh, i’m on a rich list a few times over
crustaceans are delicious
in other news the nethash is already so distributed ridiculous hardware does not makea significant dent on it
I am working on a plan where a bunch of smaller machines work together
like 32gb nodes that expose an rpc interface to read parts of the snowfield
the cool thing, is that the node does the normal 4k page reads, but can marshal a bunch of responses to send back over the network
that’ll reopen cloud mining to the fiat capable
so even though the network bandwidth is much smaller, it still uses a good chunk of the memory bandwidth
it is a pretty simple idea, if I don't do it open source, someone will do it private
but one still has to throw five figure monthly sums at this for sustaining 51% (and that seems to improve all the time)
plus the cool thing is you can just add more nodes to handle a larger field
the interims are fast enough to transfer?
interims?
pff reading code on a mobile in a forest, not yet
hah
interim reads per proof
yeah, each pow needs 96 bytes read over network
or 6 4k pages from local block device or memory
so there is a lot of benefit from batching them up and sending over network
times worst case 10M / s
ah yes, batching will make it a glacial flow sorta thing
I'm thinking LAN (or cloud virtual LAN) not internet in general
nothing is timing sensitive
for this work as long as it is less than a minute, latency doesn't matter
but bandwidth does
planning on saturating most of gigE in and out of each node
we shall see if that actually plays out
p2p within a ’cluster’ or orchestration via the pool?
for now, manually configured tightly bound cluster
mostly outgoing bandwidth requirements would ramp up pool hosting costs
maybe multicast discovery later
UDP multicast requires a well configured LAN
TCP is fine. I'll see how GRPC does for me and change it to something else as needed.
did a student campus IPTV stack on that way back when
the multicast udp would only be for discovery and configuration, not actual pow data
still, at segment borders one needs reflectors
I did some multi-router IPTV setup back in 2001 or so
no one used it
we had 5k users
nice
one can see why unicast was not an option
yeah
that is what multicast is for
Thanks for listening to my idea, I'm glad it doesn't sound too insane to you immediately
sounds like what i’ve been touting already
but some more walking to do if i’m to reach lunch in time
holy shit, it works
I have a little turd machine reading 30kh/s over the network
causing 800MB/s of read load
should be about 3 MB/s of network traffic but I don't have an easy way to graph that right now
I should really go to bed now
What do people use for traffic graphing these days?
Assuming I have a switch that supports snmp
hell, mrtg has move recent updates than I expected
i use munin for everything. it is kind of mrtg for dummies.
@fydel zabbix
munin has all i need.
but zabbix looks interesting, too.
if you hop off munin, look at ELK, TICK or splunk
Oh god, more things
I mostly get stuff done by avoiding messing with new things as much as is practical.
we need more users
true
mining snow effectively requires a level of expertise that most coins don't need
do you think getting on crappy exchanges would increase awareness/miners?
@THX 1138 4EB. No it doesn't.
Getting the highest efficiency possible requires expertise. Like everything else.
It literally just works if you merely run it.
In fact, most coins don't even work that well. We literally provide you a deployment that takes several clicks to start.
i got 5 snow over the course of two weeks or so, meanwhile people in the trade channel are selling thousands
for the amount you get, it's hard to mine and not really worth it for shit tier miners like me
i started mining snow with someone else who gave up cause they're in the same position as me
i'm still mining but it's not that attractive to miners who don't know what they're doing. i even got an nvme drive but i get more hash running my evo 750 and 850 at the same time
I'd suggest channeling all your anger into learning how to do it
People giving up is usually even more of an indicator that you should persist
But if you want to leave the hard work to a miner, you will be charged appropriately for not rolling up your sleeves
Every step of the way towards usability decreases the early and tech savvy miners edge so you still have time.
"If you can't put yourself through hell and learn linux and cloud mining solutions, i can't help you" - Satoshi Nakamoto
i could do that, and probably fail, or i could focus on my masters thesis in a completely unrelated field
and i'm not that angry; my point is that its going to put off new users
i do know of a couple of not-very-good-exchanges that snow could go on; they add anything
could do a reddit air-drop
is there a snow subreddit?
i see there is
I am working on an even more complicated miner setup if that helps
But you are absolutely correct, there is a lot to be desired in terms of ease and approachability
on the plus side its dead stable
you only mined 5 snow because there is a lot of total hashpower, and you and spending less resources than others to mine. It's not because you don't understand how to mine. Even if you were able to triple your hashrate through optimization (not very likely) you would still have just 15, still compared to those thousands.
anyways, once snow moons (which is basically an eventuality at this point) you will be able to afford all the lambos with your 5 blossoms.
So go ahead and smell the sweet nectar now!