Goal: <33% | Danger: >50%
Many know client diversity is important for a more resilient network, but they don't understand why or just how essential it is. It's not only important — it's critical. If a single client is used by 2/3rds (66%) of validators (this is currently the case) (this is currently the case) there's a very real risk this can result in disrupting the chain and monetary loss [1, 2] for node operators.
It takes 2/3rds of validators to reach finality. If a client with 66%+ of marketshare has a bug and forks to its own chain, it'll be capable of finalizing. Once the fork finalizes, the validators cannot return to the real chain without being slashed. If 66% of the chain gets slashed simultaneously, the penalty is the whole 32 ETH.
So why is >50% marketshare still dangerous? (this is currently the case)(this is currently the case) If a minority client forks, the 50%+ majority client can obtain a 66%+ majority. With no client having a marketshare over 33%, these scenarios are avoided. That's why <33% marketshare is the goal for all clients.
Execution clients are not immune. The risks mentioned above apply to both consensus clients and execution clients equally.
|* Grandine||beta||Linux, Win, macOS||-|
|Lighthouse||stable||Linux, Win, macOS, ARM|
|Lodestar||stable||Linux, Win, macOS|
|Nimbus||stable||Linux, Win, macOS, ARM|
|Prysm||stable||Linux, Win, macOS|
|Teku||stable||Linux, Win, macOS|
|Besu||stable||Linux, Win, macOS|
|Erigon (Turbo-Geth)||-||alpha & beta||Linux, Win, macOS, ARM|
|EthereumJS||alpha||Linux, Win, macOS|
|Geth||stable||Linux, Win, macOS, ARM|
|Nethermind||stable||Linux, Win, macOS|
|Silkworm||-||pre-alpha||Linux, Win, macOS|