Evaluation and Design of Performable Distributed Systems
Performability measures system performance including quality, reliability, maintainability and availability over time, regardless of faults. This is challenging for distributed systems, since the internet was designed as a best-effort network that does no
- PDF / 472,364 Bytes
- 17 Pages / 439.37 x 666.142 pts Page_size
- 110 Downloads / 222 Views
Evaluation and Design of Performable Distributed Systems Naazira B. Bhat, Dulip Madurasinghe, Ilker Ozcelik, Richard R. Brooks, Ganesh Kumar Venayagamoorthy, and Anthony Skjellum
Abstract Performability measures system performance including quality, reliability, maintainability and availability over time, regardless of faults. This is challenging for distributed systems, since the internet was designed as a best-effort network that does not guarantee that data delivery meets a certain level of quality of service. In this chapter, we explain the design, test and performability evaluation of distributed systems by utilizing adversarial components. In our approach, the system design uses adversarial logic to make the system robust. In system test, we can leverage existing, powerful attacks to verify our design by using existing denial of service (DoS) attacks to stress the system. Keywords Performability · Robust control · Game theory · Blockchain · Denial of service
9.1 Introduction Critical infrastructure is geographically distributed and vulnerable. A performable system withstands and mitigates unfavourable conditions. Distributed system components share data and information. In practice, sensor malfunctions, data communication hijacking or external disturbances can occur. The system has to tolerate disturbances in operation. To design and test these systems, we integrate adversarial logic into our approach. Game theory is the mathematics used to model conflict between
N. B. Bhat · D. Madurasinghe · R. R. Brooks (B) · G. K. Venayagamoorthy Holcombe Department of Electrical and Computer Engineering, Clemson University, Clemson, South Carolina, USA e-mail: [email protected]; [email protected] I. Ozcelik · A. Skjellum SimCenter, University of Tennessee at Chattanooga, Chattanooga, Tennessee, USA I. Ozcelik Department of Computer Engineering, Recep Tayyip Erdogan University, Rize, Turkey © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 K. B. Misra (ed.), Handbook of Advanced Performability Engineering, https://doi.org/10.1007/978-3-030-55732-4_9
211
212
N. B. Bhat et al.
rational decision-makers. In this chapter, we present a game theory-based approached for DCS design. Game analysis techniques improve performability. We model the system as a Two-person Zero-sum (TPZS) game. We introduce different disturbances and countermeasures. We start by finding the best metric(s) to measure network performance. Zero sum means that the disturbances and countermeasures have no cooperation, the success of a disturbance is equal to the lack of performance of the countermeasure, and vice versa. Two person means we consider only affects on one system and cooperation between disturbance and countermeasure is impossible. If more than one metric is used, we either use a weighted sum of the individual metrics or we could always use the value of the worst metric (this is known as the H-infinity metric). A game between these parties is established. For each component, we dete
Data Loading...
