Dependability encompasses properties such as reliability, availability, safety and integrity to ensure that computing systems operate correctly, even in the presence of faults. Achieving these properties requires careful consideration during design and throughout operation, often supported by fault tolerance techniques and system-level resilience mechanisms. These days, new computer architectures and systems are tailored for acceleration of AI applications - with high demands on performance, computational efficiency and energy efficiency. In addition, dependability-related properties can be crucial for their application.

This conference track focuses on methods and architectures that address dependability in classical computing systems, dependability of AI accelerator hardware and the application of AI for design, evaluation and run-time support of dependable systems. This includes classical approaches as well as techniques relevant to modern heterogeneous and data-intensive workloads, where AI acceleration represents one prominent example.

• Dependable computer systems, networks, embedded systems, AI accelerators
• Hard- and software-based fault-tolerant systems
• Codes for error detection and correction in memories, storage and networks
• Methods for reliability assessment and enhancement
• AI-driven and conventional failure detection and diagnosis
• Experiences on using AI for design and evaluation of dependable computer systems
• Security and robustness in data-intensive and AI-enabled Systems
• Efficient AI hardware accelerators (FPGA, ASIC, neuromorphic

All submission details incl. the page length and deadlines can be found in the Submission Guidelines.

Peter Sobe, Dresden University of Applied Science, Germany

Mahdi Taheri, Humboldt-University Berlin, Germany