Velas AIDPOS: The Consensus Algorithm That Automatically Scales Blockchain

We continue our series about AIDPoS consensus mechanism, explaining how the algorithm adapts the Velas blockchain to the current circumstances of the network, preserving throughput (processed and validated transactions) and network stability in optimal ranges.

Main Motivation (Why)

In our opinion, from the user experience perspective blockchain should always be working in instant transaction confirmation mode, just like centralized payment services, such as Visa and Mastercard. Instant transaction confirmation mode, meaning that any transaction will be processed and validated within a seconds.

Visa handles about 5000 transactions per second (although Visa claims a whopping 65,000 transactions per second, a widely debated figure). Similarly, Mastercard uses a network that claims to handle 5000 transactions per second (with no reliable, third-party source to verify). Velas has 50,000 tps.

Velas provide blockchain as a complex system that is deliberately configured for high performance. However, high performance brings with the limitation of time for synchronizing nodes. In the case of network issues or attacks on the consensus mechanism, this can lead to the dropping out of a significant number of nodes. As a result, performance can degrade greatly.

In order to prevent this, we are designing the AIDPOS algorithm which is proactively adjusting blockchain configuration to balance performance and security with preserving decentralization.

Different consensus mechanisms and approaches to balance performance and security

Overall approach (How)

The “recommender” algorithm will act as an objective “AI guardian” of the network, performing the following tasks: detection of anomalies (anomalous behaviour of nodes) and recommendation of optimal values of blockchain parameters. Below is a diagram describing the “Analysis Module” block.

General approach architecture

Analysis Module is a recommender neural system, that trained on historical data from the main Velas network and simulations, that model various possible scenarios with previously defined parameters. The module identifies and evaluates patterns and relationships across all the data sets, recommending new parameters and adjustments to increase network efficiency. This data is recorded in a smart contract and applied in the new epoch of the network. This way AI Module and blockchain work in synergy, achieving better optimization and high performance.

It should be notedn that AI will not interfere in any way in the consensus mechanism, but only optimizes the network processes, ensuring its functionality.

Main current result

During the last 3 months we have been finishing this scope of the work:

  • Data collection script (Script for Velas, Logs parser improved);
  • Simulation model (dataset for train “AI-guardian” module):
  • Launch blockchain by parameter set (can vary 89 parameters)
  • Apply load and collect performance metrics
  • Proof of concept AIDPOS. Trained model for optimization objective functions. Trained “AI-guardian” module for optimization objective functions (3 at least):
  • Performance (throughput/latency);
  • Synchronization;
  • Drop rate

Main deliverables:

The main result is the new running version of Velas blockchain with a recommendation model based on ML/DL technologies

  • Accumulating historical dataset
  • Developed data collection script;
  • Complicating simulations, adding new attack scenarios
  • Data collected from simulation processes;
  • Next version AIDPOS available on Velas Testnet
There are various benefits across the Velas userbase, from the end-user through to the Network validators and dApp developers.

Modern blockchain architectures keep many parameters constant, making them less flexible in terms of security and performance. Adaptive parameter selection can provide a good compromise between them, enabling the network to evolve dynamically. Now it’s not just a blockchain, but a blockchain with intelligence.

In the next article, we will dive into more detail on the testing stages, simulations and results we’ve achieved.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store