The Halides Model
Last updated
Last updated
Passport’s goal is to be the most performant distributed network. As a result, we lean towards pragmatism which in turn means that while our network is distributed, it is closed and permissioned.
This way we control the quality and the size of the network, allowing us to optimize for performance and cost. As a result of our goal, our network’s design and our plan to scale it are also vastly different from traditional approaches.
Traditional approaches to distributed key management rely on
Increasing the volume of nodes to be more open
Offloading most key network tasks such as authentication, communication, etc to the nodes themselves. This is in addition to the nodes storing the key shares themselves.
While this does make the network more open, it poses a couple of issues -
For any key related operations, the client has to communicate with all of the nodes in addition to the nodes communicating with each other. This slows down the performance as the size of the network grows
Our team came up with a new approach to scaling our network, called “Halides” that optimizes for performance.
Unlike the traditional model, we introduce a new component called the orchestrator and reduce the nodes’ job to managing key shares and nothing else. The orchestrator handles everything else such as authentication (via enclaves), communication and more.
Secondly the key shares in the nodes are further broken down into chunks and these chunks are clustered into what we call “halides”.
This way, the client only talks to the orchestrator which in turn is optimized to handle all other sorts of communication. And our key chunk distribution ensures that at the macro level, the volume or the size of the network is minimized while still being able to distribute smaller key chunks of the share.
While this introduces a “trust-minimized” layer called the orchestrator, it allows us to boost the overall performance of the network. It also opens up the possibility for there to be multiple orchestrator vendors in the future.
If you'd like to deepdive into the Halides model, take a look at our whitepaper (in draft) here.
