What’s staked connections and Atlas?

This phrase staked connections is new to me. What is it ? And also how do I get access to Atlas ?

this is my understanding, but it could be wrong, hopefully someone with more knowledge will weigh in here -
A node supports some number of staked/unstaked peers and a given number of connections per peer.
So if you are using a staked validator for example, you use staked connections. if you spin up your own rpc node, no stake so unstaked connection.

gm @arv_arm and @simplesimon872 !

Transactions are sent to a leader’s TPU (Transaction Processing Unit) port to be processed. This is separate from Turbine, which propagates blocks to all other validators in the cluster. The typical flow would be for a client to send a transaction to an RPC node, which forwards it to the current or next leader.

Any node can connect to the leader, but the number of connections is constrained and prioritized by stake. A given node supports some number of staked/unstaked peers and a given number of connections per peer (~8). When the max is reached, there’s a process for pruning peers/connections, and staked connections benefit from not being pruned.

With Stake-weighted Quality of Service (QoS), nodes holding more stake forward more packets. For example, if a node holds 1.5% of stake, they can send 1.5% of the packets to the leader without getting washed out.

We’ve started experimenting with Atlas and have started offering it as a service with staked connections. If that interests you, I recommend opening up a ticket on the Helius Discord and inquiring there. Note that we currently have a waitlist for this service, and it’s hard to give exact timelines since we’re constrained by the amount of stake and staked connections we have access to. In the interim, we’ve open-sourced Atlas and recommend that you try running it yourself! Even without access to staked connections, you should expect an improvement in transaction perf

Hopefully, this has cleared everything up!


1 Like