Exactly, and I strongly feel that there is a logical and well reasoned disagreement on “best practices” node architecture.
I may not have all the hard numbers figured out at my fingertips, but if you give me a few months I will…
It’s good to note that you haven’t provided any hard numbers on payout comparison on running three low reliability nodes on a single IP pipe vs. running one high reliability node running on 33% less storage space. You have presented a few finer points that I missed… but the hard numbers have not been presented.
The earnings.py script shows that for the beginning of my second month of being a node operator, I am estimated to earn $1.17 on 58 GB of Egress and $0.01 on 5.70 GB*m monthly average with a total of 65.63 GB of storage.
I’m no where near capacity on either bandwidth or storage.
I’m situated in the Northeast US, and nearly 100% of my traffic is to and from Germany. It’s unclear whether the source of the traffic is due to testing, geography, node latency, or max concurrency. And, there’s not really any method that anyone can describe which will definitely improve node performance/payout for my particular situation… since there are too many variables.
Geography plays a huge role in node traffic. And it’s nearly impossible to eliminate that variable in order to test a given node performance.
The OP question was regarding how to add more storage space to a node.
Someone suggested LVM. I strongly disagreed with that suggestion. You suggested running multiple nodes simultaneously on a single IP pipe. I disagreed with that suggestion as well… It’s good to note that the Installation documentation specifically includes a statement that running multiple nodes through a single IP will not increase payout.
In order for me to test multiple nodes running on my single IP, I would need to risk damaging losing data from the node I have running to the new nodes. This is a risk variable that you haven’t listed… There is an opportunity cost to running multiple nodes on one IP.
For anyone reading this:
I’ve argued in good faith and have presented my best knowledge at the present time.
Summary of discussion:
- Running LVM across multiple drives in order to gain more storage space should be discouraged and is definitely not included in “best practices” for any general purpose computing platform.
- Running LVM on top of RAID is perfectly fine. In this case, the LV will be included in the RAID parity and will be recoverable during drive replacement.
- It is indeterminate if a given node situated in a given geography with a given bandwidth will benefit from adding multiple nodes in single drive fashion vs. one node configured to lose 33% of space to RAID parity.
If you disagree with point #3, please provide comparison numbers for nodes you have tested showing that the payout is greater. Please also include your proximity to a satellite and your available bandwidth so that those following the discussion can make a more informed decision as to what method of node architecture best suits their particular situation.