I am creating my own private blockchain with a POA consensus. What is the difference between POA networks created by Geth (Clique) and by Parity (Aura), and are there scenarios that favor one over the other?


Since some moderator decided to delete the only answer to this question here it is again:

I do not have time to summarize it.

Good question!

I don't have an answer but I do have a link that might be worth reading

In this paper we derive the functioning of two prominent consensus algorithms for permissioned blockchains based on the PoA paradigm, namely Aura and Clique. We provide a qualitative comparison of them with respect to PBFT in terms of consistency, availability and performance, by considering a deployment over the Internet where the network is realistically modelled as eventually synchronous rather than synchronous. By applying the CAP Theorem, we claim that in this setting PoA algorithms can give up consistency for availability when considering the presence of Byzantine nodes. This can prove to be unacceptable in scenarios where the integrity of the list of transactions has to be absolutely kept (which is likely to be the actual reason why a blockchain-based solution is used). On the other hand, PBFT keeps the blockchain consistent at the cost of availability, even when the network behaves temporarily asynchronously and Byzantine nodes are present; this behaviours is much more desirable when data integrity is a priority. Despite one of the most praised advantages of PoA algorithms is their performance, our qualitative analysis shows that in terms of latency the expected loss of PBFT is bounded, and can be offset by the gain in consistency guarantees. As future work, we plan to deepen the analysis of PoA algorithms by engaging further reverse engineering tasks and thorough experimental evaluations. The final goal is to validate and possible revise our claims on the availability and consistency guarantees of PoA and PBFT, by implementing the adverse scenarios we envisioned in Section 4.1. Furthermore, we want to collect real performance measurements, both transaction latency and throughput, and to test scalability with respect to varying input transaction rates and number of nodes/authorities. Moreover, we are moving towards a formalisation of permissioned blockchains so to define a framework for benchmarking and evaluating these algorithms with a more formal approach.


  • 1
    The paper is good. It gives good insights into how Aura, Clique and PBFT work and what their strengths/limitations from the CAP perspective are. – ivicaa Oct 8 '18 at 19:24
  • I know but it got removed because I didn't summarize it. I didn't read it (skimmed it halfway) so I can't summarize it but it's still a good reference. Moderator doesn't seem to get basic things, probably butthurt over something. – Nico Vergauwen Oct 8 '18 at 19:27
  • 2
    Hi there. The problem with posting links without a summary is that the link can either be moved or become dead at any time, as has happened with numerous past answers. The moderators work according to a set of rules which generally work, and disallowing link-only answers is one of them. Here's the main site-wide Meta Stack Exchange discussion on the topic: meta.stackexchange.com/questions/8231/… – Richard Horrocks Oct 8 '18 at 19:44
  • Which is also why I explicitly stated that I do not have an answer. One reference and 0 answers is better than no references and no answers. If someone feels like summarizing the link, go ahead and I'll gladly delete my answer. – Nico Vergauwen Oct 8 '18 at 20:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.