Learning automata-based bus arbitration for shared-medium ATM switches

Area of Science:

• Computer Science
• Network Engineering
• Telecommunications

Background:

• Asynchronous Transfer Mode (ATM) remains a critical network technology despite advancements.
• Shared-medium techniques are preferred in ATM switch design for practical implementation.
• Existing arbitration schemes for shared buses in ATM switches face performance limitations.

Purpose of the Study:

• To propose and evaluate new distributed arbitration schemes for shared buses in ATM switches.
• To enhance the performance of ATM switches by optimizing bus arbitration.
• To compare the effectiveness of learning automata-based schemes against traditional methods like TDMA.

Main Methods:

• Development of four variations of a distributed arbitration scheme based on learning automata.
• Simulation and performance analysis of the proposed schemes.
• Comparison with the Time Division Multiple Access (TDMA) scheme under bursty ATM traffic conditions.

Main Results:

• The proposed learning automata-based arbitration schemes demonstrate superior performance.
• The schemes effectively utilize the bursty nature of ATM traffic for improved efficiency.
• Significant performance gains are observed compared to the TDMA scheme.

Conclusions:

• Learning automata provide an effective foundation for advanced ATM switch bus arbitration.
• The novel distributed schemes offer a promising solution for high-performance ATM networks.
• The proposed approach represents a notable advancement in ATM switch design and traffic management.