BaSE Lab_New Consensus Algorithm for Blockchain Scalability_Progress Report (3)

Ongoing
1

Title of Project

Klaytn Improvement Reserve (KIR): New Consensus Algorithm for Blockchain Scalability (BaSE Lab)

Project Duration

  • Proposed Period: 2021.09.08 ~ 2023.02.28
  • First Milestone: 2021.09.08 ~ 2022.02.28
  • Second Milestone: 2022.03.01 ~ 2022.08.31 (Current)
  • Proposed Third Milestone: 2022.09.01 ~ 2023.02.28

Goal and Needs of the Project

Research and Development Goals

  • The limitations of the consensus algorithm used by Istanbul Byzantine Fault Tolerance (IBFT) used by Clayton are proposed to improve the performance of the Clayton platform through the “A Coordination Technique for Improvement Scalability of Byzantine Fault-Tolerant Consensus” (Applied Sciences, Vol. 10, 2020), a study held by Sogang University’s Intelligent Blockchain Research Center

Necessity of Research and Development

  • The consensus algorithm of the blockchain network allows all nodes to agree on which transactions are valid and should be added to the distributed ledger. In addition, the consensus algorithm ensures that all nodes can consistently see the current state of the distributed ledger even when faced with network interruption or attack.

  • Consensus algorithms play an important role in maintaining the security and integrity of blockchain networks, but can also affect network speed.

The Result of Third Milestone

  • Consensus algorithm consensus process speed measurement

    • IBFT consensus algorithm and coordinator consensus algorithm operation speed measurement
    • IBFT sum speed: 8 ms average, Coordinator sum speed: 4 ms average
      • approximately 50 percent improvement has been identified

  • Network TPS Measurement Experiment with Physical Node Construction

    • No significant difference was found when compared to the secondary experiment in which the virtual node was constructed.
      • IBFT consensus TPS test result: 3rd average TPS: 16th maximum TPS: 33.16
      • Coordinator consensus TPS test results: 3rd average TPS: 15.88 3rd maximum TPS: 32.68

  • In conclusion, even in a physical environment, the coordinator technique did not produce better results than the IBFT consensus method, but it was confirmed that the performance was maintained the same as in a virtual environment.

  • Progress of load test

    • Conduct load tests to identify the environment where the blockchain network is abnormally terminated.
      • Continue to increase the number of nodes and check when the blockchain network will be terminated.
      • Experiment with continuous virtual node growth on one physical PC

  • As a result of the experiment, in an environment with 48 nodes, IBFT generates an average of 550 blocks and the network is forcibly shut down.

  • However, when the coordinator technique is applied, an average of 1050 blocks are created in an environment with 48 nodes and the network is forcibly shut down.

  • Compared to Clayton’s IBFT consensus algorithm, it was found that the load resistance of the blockchain network was high when the coordinator technique was applied.

Project Goal

The goal is to improve Klaytn platform’s scalability, which has limitations of IBFT (Istanbul Byzantine Fault Tolerance), to make the performance of consensus algorithm better by applying a research “A Coordination Technique for Improving Scalability of Byzantine Fault-Tolerant Consensus” (Applied Sciences, Vol. 10, No. 21, 2020) that Intelligent Blockchain Research Center at Sogang University had conducted.

Advantages

  • The coordinator algorithm presented by Second Milestone needed to improve decentralization because the coordinator reached an agreement with a fixed state. Decentralization was confirmed by randomizing the coordinator election by applying the round robin function.

  • In order to configure the experimental environment variables similar to the physical environment in which Clayton is actually driven, an experiment is conducted by configuring a physical node, and the result value is confirmed.

  • In order to test the load in an environment where the number of nodes increases, the number of virtual nodes was set to 48 and the test was conducted in an environment with high complexity, and it was confirmed that the coordinator agreement method was more load resistant than IBFT.

Conclusions

  • As a result of comparing the IBFT consensus rate and the coordinator agreement rate, it was found that the coordination agreement algorithm was superior to the IBFT agreement algorithm(improvement result: 8ms→4ms).

  • In a situation where the number of nodes is increasing, it is found that the coordination agreement algorithm has higher load resistance than the IBFT agreement algorithm. → Reduced network memory consumption.

  • Omitting one of the processes of the agreement. Compared to the IBFT consensus algorithm, the coordinator consensus algorithm is faster, which leads to increased load resistance in an environment with increased complexity as the number of nodes increases.

Full Report: 3rd Klaytn Report - Google Docs