Goals
Public cloud platforms are rapidly becoming the foundation of our Information-based economy. Cloud providers, which already offer a wide of diversity of services, continue to expand their offerings. These services range from mature and well-established Infrastructure-as-a-Service (IaaS) (e.g., bare metal servers, virtual machines, or containers) to various Platform- and Software-as-a-Service (PaaS/SaaS) offerings to the more recent Function-as-a-Service (FaaS). Users are adopting cloud platforms due to both their quantifiable cost savings over using in-house infrastructure, and their ease-of-use, which improves productivity due to "toil reduction" and "distraction reduction." While security and data privacy remain a concern, recent advancements have mitigated some of these concerns and also demonstrated some advantages of cloud platforms over in-house infrastructure.
Cloud providers' primary focus thus has been on offering a diverse suite of resources and services to attract demand and capture a large user-base. However, growing competition will make it increasingly important to operate their infrastructure more efficiently (e.g., by leveraging yield management techniques to multiplex multiple diverse workloads and even overbook resources). In general, well-functioning markets are important for maximizing resource efficiency, as markets bring together buyers and sellers with the goal of ensuring all resources are sold at an equilibrium price that balances supply and demand. As public cloud platforms have sought to attract demand and increase their utilization, they have naturally evolved to resemble nascent markets for IT resources.
This evolution raises new research challenges for both users and providers that are inherently economic. Users must design cost-aware systems that are capable of navigating the cloud market's complexity by automatically selecting and configuring cloud services to balance a range of metrics, including ease-of-use, cost, risk, performance, availability, etc. Providers, meanwhile, must design and offer cloud services that are easy to use, maximize utilization and revenue, satisfy diverse application requirements, and cannot be gamed. While the computer systems and networking communities have a long history of research into various forms of economic resource allocation, the emerging economics of the cloud differs from prior research in multiple important respects, as summarized below.
- Explicit Costs. Whereas many prior systems and networks have employed market-based mechanisms for resource allocation, these markets and their associated currencies were often virtual. Public cloud platforms charge real money and thus make the cost of using resources and services explicit.
- Complex Choices. Cloud platforms offer myriad services to choose from with different abstractions at multiple levels of the software stack with distinct pricing models and tradeoffs. The abstractions vary widely, from the more conventional (virtual) machines to the more recent software interfaces (SaaS and FaaS).
- Risk Exposure. Cloud platforms expose applications to many novel forms of uncertainty and risk, both explicit and implicit, that applications must consider when selecting services, as these risks have implicit costs.
- Increased Dynamism. Cloud platforms are arguably an even more dynamic environment than traditional systems, often with prices and/or resource allotments that vary over time based on the collective application demands and workload intensity.
- Design Impact. Optimizing for the criteria above affects the design of cloud systems and applications, since these systems have an incentive to respond and adapt to workload and price dynamics by continuously re-configuring themselves. In particular, many issues related to "mechanism design" (e.g., choosing pricing structures, choosing performance SLA levels) are fundamentally novel and may not be mere adaptations of existing solutions from other domains.
The goal of this workshop is to bring together researchers from computer systems and networking and from the fields of economics and operations research to define the important research challenges at this intersection over the next 5-10 years. A key goal is to understand how services offered by cloud platforms affects the design of cloud systems and applications, which have historically optimized for traditional metrics, such as performance, availability, energy-efficiency, etc., and not economic metrics, such as cost and risk. In particular, the objectives and intellectual merit of the workshop include:
- Foster a community working at the intersection of cloud computing and economics that includes academic researchers from computer systems design and economics, finance, and operations research.
- Set the vision for this community and identify the key challenges and open problems related to designing services for cloud markets and designing cloud applications to optimize for those markets.
- Engage industry participants that are actively working at this intersection, introduce them to this community, and get their feedback on key challenges and open problems from their perspective.