What are the hardware specs to run an RPC?
Solana RPCs are nothing to mess with. They do almost all of the things that validator nodes do PLUS they handle almost all of the lookup requests. Ever opened your Phantom wallet and waited for the balances to load? That's because it was blowing up an RPC requesting all of the balances and SPL tokens and NFTs in your wallet.
Now imagine that happening for all the wallets everywhere in the world. That's an RPC, and that's why your desktop PC will commit seppuku if you try to run a Solana RPC on it.
- More important than anything is redundancy. You run servers in data centers because you get:
- Dual power circuits from 2 separate power companies
- Dual battery backups
- Dual ISPs
- Dual cooling and air conditioning flows
You don't have those things at home, so just know going into this that if someone is paying you for a - and this is a keyword here - *Premium* service such as *Premium* RPC, you have an obligation to provide just that.
So what data center should you use and where to begin this journey? 👇
Apply for access to data centers through the Solana Server Program (requires KYC!) https://solana.org/server-program
You can alternatively explore other bare metal providers without going through Solana Server program, like Latitude.
Once you've gone through that, you will be given a catalog of servers and data centers to pick from. I'm here to tell you right now, most of those offerings work great for validators - they do NOT work great for RPCs. Never forget, RPCs require a hoss of a server.
- AMD EPYC 7502P 32 Core CPU or better (AMD EPYC is preferred over Intel, and the 7443 will work)
- 512 GB RAM minimum, 1 TB RAM preferred
- Disk layout:
- 1x OS Drive that will also hold Solana logs. 128GB min, 256GB+ preferred
- 2x 1TB nVME drive minimum, 2x 3.9TB nVME preferred
- And another thing to watch out for is bandwith usage. Depending on the region, some of our nodes can transmit up to 10 TB outbound per day, which can rack up a big bill if your hosting provider charge by the TB.
The preferred node that most operators have been using is currently the Equinix EQ-6 or Latitude s3.large.
NOTE: IN ADDITION to the hardware requirements above, it is also required that all operators stake 10,000 SHDW per node that they want to operate!!