I will provide a historical perspective on wait-free simulations of multi-bit shared registers using single-bit shared registers, starting with classical results from the last century and ending with an overview of the recent resurgence of interest in the topic. Particular emphasis will be placed on the space and step complexities of such simulations.

Significant paradigm shifts are occurring in the way data is accessed and updated. Data is “very big” and distributed across the globe. Access patterns are widely dispersed and large scale analysis requires real- time responses. Many of the fundamental challenges have been studied and explored by both the distributed systems and the database communities for decades. However, the current changing and scalable setting often requires a rethinking of basic assumptions and premises. The rise of the cloud computing paradigm with its global reach has resulted in novel approaches to integrate traditional concepts in novel guises to solve fault-tolerance and scalability challenges. This is especially the case when users require real-time global access. Exploiting edge cloud resources becomes critical for improved performance, which requires a reevaluation of many paradigms, even for a traditional problem like caching. The need for transparency and accessibility has led to innovative ways for managing large scale replicated logs and ledgers, giving rise to blockchains and their many applications. In this talk we will be explore some of these new trends while emphasizing the novel challenges they raise from both distributed systems as well as database points of view. We will propose a unifying framework for traditional consensus and commitment protocols, and discuss novel protocols that exploit edge computing resources to enhance performance. We will highlight the advantages and discuss the limitations of blockchains. Our overall goal is to explore approaches that unite and exploit many of the significant efforts made in distributed systems and databases to address the novel and pressing needs of today’s global computing infrastructure.

Distributed systems are constantly faced with difficult decisions to make, such as in scheduling, caching, and traffic routing to name a few. In most of these scenarios, the optimal decision is unknown and depends heavily on context. How can a system designer know if they have deployed the best decision-making policy, or if a different policy would perform better? As a community, we have developed a few methodologies for answering this question, some of them offline (e.g., simulation, trace-driven modeling) and some of them online (e.g., A/B testing). Neither approach is satisfactory: the offline methods suffer from bias and rely heavily on domain knowledge, and the online methods are costly and difficult to deploy. What system designers ideally seek is the ability to ask “what if” questions about a policy without ever deploying it, or counterfactual evaluation. In this talk, I will show how reinforcement learning and causal inference can be synthesized to counterfactually evaluate a distributed system. We will apply this methodology to production infrastructure systems in Azure, and face fundamental challenges and opportunities along the way. This talk will serve as an introduction to reinforcement learning and the counterfactual way of thinking, which I hope will interest and inspire the OPODIS community.