Reducing Computational Overhead

At scale, costs increase substantially due to several factors, including higher memory demands for state storage, increased network bandwidth for write replication and state snapshot distribution, greater computational requirements for data processing, and the potential for longer recovery times in case of failures.

Separating computation from consensus leads to increased throughput, scalability and efficiency. With a little bit of trust in “trusted computing” there are significant gains to be achieved.

• Off-chain storage nodes store ledger states and simulate smart contract execution

• On-chain consensus nodes maintain only the short commitments of ledger states

The computation layer does not require their entire group of nodes to agree on results. Usually a smaller portion of the nodes(committee) executing the same output is enough for it to be assumed those results are correct.

Instead of having 100 validators all running the same program using consensus to agree on execution results there could instead be 10 compute nodes performing the execution, each creating a proof for it, and then 90 other validators running that proof as their consensus to apply the state changes to their ledger.

Key Idea and Distinction!

We would like heterogeneity to ensure that the outputs do in fact match even when executed in different TEE systems. Furthermore, multiple enclaves should be used to evaluate if the same input parameters to different manufactured TEE’s will output the same result. This negates the possibility that a vulnerability introduced by the manufacturer will upend the network, getting us closer and closer to trusting the results by virtue of the diversity of hardware environments we can apply to the execution.