How do Tps scale with decentralization of stake?

Hi, correct me if I’m wrong here, but it seems to me that a lot of Solana’s speed is due to exploitation of centralization in the network. The leader authority is assigned to nodes randomly weighted by stake, which means the the highest staking nodes are leading the network most of the time. Now I don’t know how exactly everything works, but I imagine that reactivating a previous leader gives a lot of opportunity for optimization, by caching for example. Also higher staking leaders tend to be professional computing centers with better hardware.

Now I am wondering how the Solana Tps would scale when the network becomes more decentralized. I understand that it is not meant to be as decentralized as Iota for example, which is designed for “Internet of things” i.e. favors as many nodes as possible. But how bad is it really? Are there any benchmarks that have been run on larger networks with highly/evenly decentralized stake?

That just isn’t true, all validating nodes process the full blockchain state, and they all have to validate it no matter who the leader is.
It doesn’t take less effort for a previous leader to return, vs a different validator who hasn’t been leader for a longer time. I don’t think there’s any kind of data that could be “cached” in that case to make it faster.
The requirements for running a node are similar for all validators.

Solana scales by being efficient and having high hardware requirements for validators. Of course it will never have as many fully validating nodes as a much lighter cryptocurrency like Bitcoin.
Running a Solana node is far more expensive. You can’t do it at home with a spare computer, you need to rent a server with symmetric gigabit bandwidth at least, as well as hardware equivalent to a current generation gaming desktop computer. That’s a few hundred dollars per month, so it’s not that unreasonable, many people can afford it, especially cryptocurrency investors.

And you make money from running such a node, since you receive a certain percentage of the earnings from people who are staking with you. So it could even be profitable.

Solana is already running with hundreds of such validator nodes active.

That’s still not extremely decentralized and I don’t believe we will ever reach a network of Bitcoin’s size. But it’s not centralized either, and it will get better since hardware gets faster and cheaper over time. Solana is currently capable of 60000 tx/s. As long as the developers don’t raise the requirements to an unreasonable degree in the name of higher throughput, and try to make do with the already ample blockchain capacity, we will see the cost of running a validator decrease over time.

All of the validators are fully participating in the network regardless of the % of total stake they hold, and assuming they are working properly, there should be no issues even if the stake is evenly distributed among them. The problem right now is that it’s not, most of the stake is in a few large validators, but they’re trying to fix this problem.

The main issue I have with the current system is that, even though there are measures to limit this in place (target signatures per block and such), validators have an incentive to increase the number of transactions per second that are processed, because they make money from the transaction fees.
That raises hardware requirements, and can be a force to increase centralization over time.

1 Like