IOTA Research Status Update December 2019
The full article was originally published by Serguei Popov on Medium. Read the full article here.
This month’s update is about the status of our major task, Coordicide.
Our team is split into several subgroups, each of which is focused on a specific area of Coordicide. The DRIs (Directly Responsible Individuals) for each group meet regularly to discuss developments and impacts across the groups. We have found this workflow to be very effective.
Our primary goal is to deliver a specification for the Bee node software, which will form our implementation of Coordicide. This will run first on a testnet, and then later on the Mainnet. Our team is currently engaged in three tasks: writing a research specification, whitepaper revisions and researching the Coordicide components.
This specification is similar to a whitepaper, but with sufficient technical detail for it to be transformed into our next deliverable: the engineering specification. The engineering specification will then be handed over to our developers to code, test, and release.
We have received wonderful feedback from the public and our academic collaborators, including members of the IOTA Research Council. We are now writing a revised whitepaper in light of our research. This paper is extremely important to us, forming the base through which experts can verify the core of our solution.
Coordicide component research
Coordicide is made up of several components. While we are actively researching numerous topics, our focus in this post is the major Coordicide components necessary for completing the Research specification. Below we provide status updates on these.
We have two proposals for autopeering: “salt-based” and “arrow.” We have been analyzing the salt-based proposal both analytically and through simulations in GoShimmer. In particular, we have been investigating how long connections between nodes will last, and also studying the strength of the “verifiable randomness” test. We have also started modeling the fluctuations on the number of connections of the nodes’ graph. Soon we will compare the simulation and analytic results and compile them into a number of scientific publications. Once complete, we will be able to finalize autopeering the spec for handover to the engineering team.
Fast Probabilistic Consensus (FPC)
We have made significant progress with Berserk detection, and a blog post is forthcoming. The blog post is based on the IOTA.cafe thread on the topic. We will discuss a safety feature that makes it impossible to be berserk in FPC. By implementing those features we will decrease the efficacy of possible attacks, leading to decreased probabilities of agreement and termination failures. This is confirmed by both the FPC paper and our simulations.
The actual implementation of FPC will depend on parameters defined by the reputation system and mana of the nodes. Given that the precise distribution of the mana post-coordicide is variable and unknown, our approach is to study the impact of various mana distributions on the robustness and efficiency of FPC. This analysis will allow us to determine the final parameters of FPC, and complete the specification. Completing this analysis is the main focus the FPC group.
The autopeering network protocols are designed so that Cellular Automata can achieve fast and robust convergence. However, the existing theory does not cover all practical situations. To integrate CA, we first determined whether the associated analytical tests provided the desired convergence. These initial studies reinforced what we expected from the outset — that it would be difficult to get new theoretical results for CA.
The logical next step was then to perform analysis through simulation. For this purpose, CA has been implemented in GoShimmer: https://github.com/iotaledger/goshimmer/tree/ca. We will use this implementation to run tests and simulations. This will allow us to check if the expected robustness holds.
As a voting protocol, CA distinguishes itself by having extremely fast convergence, and it is thus suitable for IoT applications, where we expect it to shine. FPC provides a more robust convergence. Based on the results from the simulations, we will determine whether to include CA in the first version of Coordicide. We expect both to coexist in future optimizations.
We are currently extending our previous paper on adaptive Proof of Work, which we will submit to the IEEE ICBC Conference as a follow up of last year’s publication. After that, we plan to integrate the adaptive Proof of Work into GoShimmer for testing and parameter tuning.
In parallel, we are investigating the use of a Verifiable Delay Function (VDF) as an anti-spam mechanism. The research focuses on generating a public RSA modulus in a distributed way, which will be used in a Wesolowski-like VDF. In November, we submitted two work in progress papers to the Stanford Blockchain Conference. We are now writing a third one on how to integrate VDFs as a rate control mechanism in IOTA, which we plan to submit to ACM MobiHoc.
Additionally, we are working on the definition of a congestion control mechanism for the Tangle. The main challenge is to guarantee that the nodes share the same vision on incoming transactions (or, equivalently, preventing network splits) by efficiently using the network bandwidth. We have sketched a protocol proposal inspired by the Additive Increase Multiplicative Decrease algorithm commonly used for the Internet, and we are currently building a simulator to evaluate its efficiency. After this validation phase, the next task is to provide full specifications for Coordicide.
Various packages have been integrated on the hive.go repository. Common packages and functionalities between GoShimmer and Hornet are organized together under this single repo. The new salt-based autopeering module (from autopeering-sim) has been integrated into GoShimmer and is currently in testing/debugging phase. The ledger state is near completion, currently in the testing/debugging phase. The gossip layer has also been refactored.
These updates to GoShimmer are essential because they make the overall implementation more robust and reliable. This, in turn, will allow us to focus on integrating the voting layer.
The pieces of Coordicide are coming together! Please stay tuned for further updates, and as always, feel free to reach out in our regular communications channels. You can stay up to date with the IOTA Research team in the #tanglemath channel on our Discord. We have also recently launched our public forum: IOTA.cafe. Please do check out these resources for further details on what we have outlined above.