Hi there,
I want to create an Algorand Node (non-relay and non-archival) with an AWC EC2 instance.
There are any recommendations about the EC2 instance type?
In order to estimate bandwith/transfer usage, does exists an “estimated” number of transfer data used to run a node monthly (obviously it change from to month, but some estimation would be nice)?
Thanks in forward!
On my personal node (on Linode), I see around 500GB ingress and 20GB egress every month. (If you’re participating with high stakes and proposing many blocks egress may be higher though.)
If you participate, it is very important to use SSD (EBS SSD should be good enough) to be able to vote on time and participate correctly.
If your node is not a participation node and is not used for some critical dApp, you may be able to use cheaper disk (for example on GCP, balanced persistent disk is good enough). The cheaper disk may not absorb as well peak traffic but in the worst case, your node gets a bit delayed (or crash) but that would have no effect on the health of the network
The node requirements in https://algorand.foundation/algorand-protocol/network have now been updated and are what we usually recommend.
If you are participating with a small stake (below 1M Algos) or not participating at all, you can definitely try to start with 2vCPU, 4GB RAM and see if it works or not and increase if necessary.
If you are participating with a larger stake, it is recommended to start with higher specs.
In general, the most critical part of an Algorand node is the speed of the disk and the quantity of RAM. The number of vCPU is rarely an issue from my experience.