Cloud@MICRO Virtual Workshop

Overview

Over the past decade, Cloud Computing has transformed the computing industry. Economies of scale, financial and technical advantages of Infrastructure as a Service (IaaS), and the rise of mobile computing have combined to make Cloud Computing popular for an increasingly broad set of workloads. Remote work and the pandemic further accelerated these powerful industry trends. Many of the largest installed computing resources nowadays are now run by Cloud Hyperscalers.

There has been little reported work on investigating requirements of Cloud hardware or Cloud system design. The original Cloud value proposition of deploying efficient ways to use general-purpose (e.g., commodity) servers rather than customized or specialized hardware might be the reason behind the limited reported research on Cloud hardware. While use of general-purpose servers may have been true at the outset of Cloud Computing, the Cloud usage model has been evolving from single tenant bare-metal servers and VMs to multi-tenant systems using containers and now to microservices. Each usage model has different implications and requirements on the design of compute, network, and storage components. Multi-tenancy usage models demand zero-trust architectures that elevate security and isolation as first-class design constraints. Meanwhile, the slowing of Moore's Law and the dramatic growth of compute intensive AI workloads and Cloud-based HPC have spurred innovation in accelerators and offloading compute from the CPUs. As a consequence, systems requirements and design needs for Cloud Computing are becoming increasingly different from the requirements that have driven the development of traditional servers by hardware vendors, so that Cloud Hyperscalers are increasingly customizing and building their own systems and devices across the spectrum of compute, acceleration, network, storage, and security. Cloud architectures are adopting, and in some cases pioneering, a variety of compute devices for better performance and workload isolation. The future of Cloud and IaaS is evolving rapidly, in particular when considering the introduction of potentially transformative technologies on the horizon, such as Quantum Computing.

This workshop is intended to bring together researchers, developers and practitioners from academia and industry, to share experiences and implementations related to the architecture, design and implementation of Cloud systems infrastructure, and encourage discussion of future trends and research in these areas. The workshop, held virtually, will include invited presentations by experts from Cloud providers, as well as an open call for participation. Contributions, in the form of a presentation or paper plus presentation, will be solicited in the areas listed below, albeit not limited to them.

Topics of interest

We are soliciting original contributions that address a wide range of theoretical and practical issues including, but not limited to:

• Architecture of systems and components optimized for Cloud use cases such as VMs, containers, microservices, offloading of management and agents, etc.
• Reliability and availability in Cloud systems
• Hardware-based security for Cloud systems
• System composability and hardware disaggregation in Cloud systems
• Use cases and examples of hardware customized for Cloud, emphasizing the criteria for needing customized hardware
• Future of Cloud infrastructure, from a hardware perspective: what is different, what stays the same
• Network and storage architectures for massive multi-tenancy needed for Cloud systems