So far no info about the ram amount per node. V2 used 4 gigs per core (in my case g4560 + 16 gigs ram ) and that was fine. If V3 will use less … awesome
Pi’s with 1gb run into memory issues from time to time, but I’ve never heard of issues with 2gb setups. So I recommend 2gb per node.
To be honest, anything over 1gb per node is probably fine. I think you’d be perfectly fine running 3 nodes with 4gb on 3 different hdds. Please note that running more than 1 node is only useful if you want to share multiple hdds. There is no advantage over running a single node in terms of data or traffic you receive. They will be treated as a single node.
I agree with @BrightSilence absolutely min. is 1,5GB (without issues), would be better 2GB (min.) or more especially more RAM makes sense if you have slow disk system (RAM will be like a buffer)
I know. But more real ram usage will be seen some time later on final V3 release. Now i have a 8core/32gb rig ready for that. if will be less - awesome
I actually don’t expect the RAM requirement to go up by much. These learnings were based on artificial high load tests during early testing with an artificially low number of nodes. The more nodes there are, the more peak usage will be spread across nodes and the lower the RAM demand.
Additionally it’s worth to keep in mind that when running multiple nodes on 1 system, the traffic is spread across those nodes anyway, so apart from some RAM overhead for duplication of core processes, I would actually not expect a large RAM increase for running multiple nodes. And you may even be able to get away with running 4 nodes on 2GB.
That last part is speculation though and not yet based on actual tests or user feedback. I am however pretty confident in saying you will never need anywhere near 32GB, no matter how many nodes you run. You’ll run into other bottlenecks way before the RAM is an issue.
Multiple nodes on one IP is allowed now. Keep in mind though that other than being able to share multiple drives there are no advantages to that setup.
I recommend running a node per drive, but it’s probably not a good idea to start them all up at the same time. Nodes go into a vetting phase first which requires 100 successful audits on each satellite. If your traffic and audits are spread out over say 4 nodes, it will take 4x as long to get vetted.
Best approach in my opinion is to just start with the first node (largest hdd) and spin up the second once the first one is about 75% filled and so on.
FYI you can now request multiple auth tokens with the same email address. But, you need to use each to sign an identity before you request the next auth token, otherwise the same unclaimed token will again be dispensed. Also, only one token per day is the current limit to prevent spamming. In any case, even before, you could have used your regular email address in all your nodes´ configurations, there was never a requirement to use different addresses there before, just for requesting the tokens.
It is not advisable to start multiple nodes at once to begin with, they will share traffic for the same total as a single node would and all nodes need to be vetted which will only take longer if you do it all at the same time.
I advise starting your first node, wait until it’s about 75% full, then spin up the next and repeat that process. Unless you have very small hdds I don’t expect you will be using all 8 any time soon using that method.