Overview

This ordering service has a similar architecture to the Kafka-based ordering service already provided by HLF. It is comprised by a set of 3f+1 ordering nodes and an arbitrary number of frontends, as depicted in the figure bellow.

The ordering nodes are equivalent to the Kafka-cluster with a Zookeeper ensemble, in the sense that they also execute a distributed consensus protocol responsible for establishing a total order on transactions. Furthermore, the frontends are equivalent to the ordering service nodes (OSNs) used by the Kafka-based ordering service, in the sense that they also relay the transactions issued by clients into the the consensus protocol. However, some key differences between this service and the Kafka-based service are:

1. The ordering nodes execute a BFT consensus protocol, which means malicious nodes are unable to disrupt the service (as long as they do not exceed f nodes out of a total of 3f+1);
2. Whereas Kafka's OSNs receive a stream of ordered transactions, this service's frontends receive a stream of pre-generated blocks containing ECDSA signatures from 2f+1 ordering nodes;
3. Whereas Kafka's OSNs are logically comprised by a single Go process, this service's frontends are comprised by two processes (a Java and a Go component).

For more information regarding this project, check out the technical report available here