We have deployed solana validator using aws blockchain node runners, based on baserpc configuration, but we changed from r6a.8xlarge to r6a.16xlarge and increased data volume and accounts volume to 16 0000 iops and 1 000 throughput on gp3 volumes.(data 2TB and account vol 1.5TB) root volume 100gb, iops 5000, throughput 250 Running on solana version 1.18.22, aws region us east 1 Our node caught up with the network during the first 24 hours, but now we are seeing slots behind rose to around 200 and jumps up and down again, and for a bit stays in 200 range. In cloudwatch dashboard we are not seeing any bottlenecks, or probably we are not looking in correct place,If someone has any idea - any guidance will be appreciated.
1 Answer
To run a Solana validator or RPC, you definitely need a bare metal machine. While it may be advertised, using AWS or GCP is not recommended (though it may be possible).
The typical bottlenecks are disk IOPS and CPU.
16K IOPS sounds appropriate, but I’m curious if that’s the actual performance you’re getting. You could run some tests using FIO or another disk benchmarking tool.
Regarding the CPU, EC2 vCPUs don’t compare well to physical CPUs. Running a Solana validator involves significant context switching, and virtual machines are typically slower.
As a tip, there isn’t much documentation available online, so you’ll often need to dig into the Solana Labs Discord. The community is helpful, so feel free to ask for help there.
