КОНФЕРЕНЦІЇ ВНТУ електронні наукові видання, 
Молодь в науці: дослідження, проблеми, перспективи (МН-2026)

Розмір шрифта: 
REAL-TIME PERFORMANCE MANAGEMENT OF PRIVATE SOLANA DEPLOYMENTS USING MODEL PREDICTIVE CONTROL
Олександр Мирославович Хошаба

Остання редакція: 2026-02-18

Анотація


Private deployments of high-throughput blockchain infrastructure increasingly serve as internal transaction and audit substrates for enterprise systems, where predictable confirmation time and bounded tail behaviour under bursty load are often more valuable than peak throughput. This paper presents an engineering concept for performance management of a private Solana cluster using an adaptive control loop implemented at the ingress gateway. The approach treats the end-to-end transaction processing path as a macro-scale queueing process and regulates user-perceived performance using Model Predictive Control (MPC) in a Multi-Input Multi-Output (MIMO) setting. Control inputs are selected to be practical in real deployments without intrusive validator instrumentation and include gateway admission shaping and a priority-fee policy based on compute-budget parameters, while outputs include high-percentile confirmation latency, median confirmation latency, confirmed throughput expressed as transactions per second, and a success ratio reflecting timely attainment of the target commitment status. Because effective service capacity and contention patterns vary over time, the method incorporates online system identification using a lightweight recursive estimator with persistent excitation procedures suitable for commissioning windows. The receding-horizon optimiser enforces service-level objectives for tail latency and reliability through soft constraints and introduces an explicit cost proxy to discourage economically inefficient “fee escalation” strategies. As production metrics are not yet available, the paper defines an evaluation plan based on scenario-driven workload regimes (baseline, excitation, stress), prediction-consistency checks, and operational acceptance criteria expressed through violation frequency of tail-latency objectives, throughput stability, and controlled fee expenditure. The proposed design provides a practical bridge between control engineering and blockchain operations for private Solana deployments, enabling systematic tuning and measurable service governance with lightweight measurements.REAL-TIME PERFORMANCE MANAGEMENT OF PRIVATE SOLANA DEPLOYMENTS USING MODEL PREDICTIVE CONTROL

Ключові слова


private Solana cluster; performance management; model predictive control; multi-input multi-output control; tail latency; transaction confirmation; gateway admission control; online system identification; queueing-based modelling; priority fees

Посилання


1. Zou D., Lu W., Zhu Z., Lu X., Zhou J., Wang X., Liu K., Wang K., Sun R., Wang H. OptScaler: A Collaborative Framework for Robust Autoscaling in the Cloud [Electronic resource] // Proceedings of the VLDB Endowment (PVLDB). - 2024. - Vol. 17, No. 12. - P. 4090–4103. - DOI: 10.14778/3685800.3685829. - URL: [https://doi.org/10.14778/3685800.3685829](https://doi.org/10.14778/3685800.3685829) (accessed: 27.01.2026).

2. Solana Foundation. Transaction Fees [Electronic resource]. - URL: [https://solana.com/docs/core/fees](https://solana.com/docs/core/fees) (accessed: 27.01.2026).

3. Anza. Solana Commitment Status [Electronic resource]. - URL: [https://docs.anza.xyz/consensus/commitments/](https://docs.anza.xyz/consensus/commitments/) (accessed: 27.01.2026).

4. Solana Foundation. Transaction Confirmation & Expiration [Electronic resource]. - URL: [https://solana.com/developers/guides/advanced/confirmation](https://solana.com/developers/guides/advanced/confirmation) (accessed: 27.01.2026).

5. Solana Foundation. Network Performance Report: July 2023 [Electronic resource]. - URL: [https://solana.com/news/network-performance-report-july-2023](https://solana.com/news/network-performance-report-july-2023) (accessed: 27.01.2026).

6. Alizadeh S., Khabbazian M. Solana’s transaction network: analysis, insights, and comparison [Electronic resource] // EPJ Data Science. - 2025. - Vol. 14. - Article 48. - DOI: 10.1140/epjds/s13688-025-00561-x. - URL: [https://doi.org/10.1140/epjds/s13688-025-00561-x](https://doi.org/10.1140/epjds/s13688-025-00561-x) (accessed: 27.01.2026).

7. Diamandis T., Evans A., Chitra T., Angeris G. Designing Multidimensional Blockchain Fee Markets [Electronic resource] // 5th Conference on Advances in Financial Technologies (AFT 2023). Leibniz International Proceedings in Informatics (LIPIcs). - 2023. - Vol. 282. - P. 4:1–4:23. - DOI: 10.4230/LIPIcs.AFT.2023.4. - URL: [https://doi.org/10.4230/LIPIcs.AFT.2023.4](https://doi.org/10.4230/LIPIcs.AFT.2023.4) (accessed: 27.01.2026).

8. Roughgarden T. Transaction Fee Mechanism Design [Electronic resource]. - arXiv preprint, 2021. - arXiv:2106.01340. - URL: [https://arxiv.org/abs/2106.01340](https://arxiv.org/abs/2106.01340) (accessed: 27.01.2026).

9. Rawlings J. B., Mayne D. Q., Diehl M. Model Predictive Control: Theory, Computation, and Design. - 2nd ed. - [Electronic resource]. - Nob Hill Publishing, 2022. - URL: [https://sites.engineering.ucsb.edu/~jbraw/mpc/](https://sites.engineering.ucsb.edu/~jbraw/mpc/) (accessed: 27.01.2026).

10. Ma L., Liu Z., Xiong J., Jiang D. QWin: Enforcing Tail Latency Service Level Objectives at Shared Storage Backend [Electronic resource]. - arXiv preprint, 2021. - arXiv:2106.09206. - URL: [https://arxiv.org/abs/2106.09206](https://arxiv.org/abs/2106.09206) (accessed: 27.01.2026).

11. Alom I., Ferdous M. S., Chowdhury M. J. M. BlockMeter: An Application Agnostic Performance Measurement Framework for Private Blockchain Platforms [Electronic resource]. - arXiv preprint, 2022. - arXiv:2202.05629. - URL: [https://arxiv.org/abs/2202.05629](https://arxiv.org/abs/2202.05629) (accessed: 27.01.2026).


Повний текст: PDF