University of Newcastle, UK
Benchmarks and Models for Blockchain: The Incentives Layer
We consider blockchains from a performance engineering perspective, with an emphasis on the incentives layer. The incentives layer in the blockchain software stack refers to the mechanisms, protocols and software architecture associated with fees, costs, and other considerations that either entice or discourage participants. The presentation builds on two earlier keynotes that consider blockchain performance engineering in general and in the consensus layer, respectively. We start the presentation with an illustration that incentives need to be aligned to ensure the reliable operation of permissionless blockchains. This forms the motivation behind a list of main topics that require further attention from the research community. We will distinguish two categories of issues at the incentives layer: issues related to the dependable operation of blockchains in general, and issues related to incorporating incentives in novel blockchain applications. With respect to the first category, we report on an extensive benchmarking study of Ethereum smart contracts, which explores the relation between the rewarded fee and the computational cost. With respect to the second category, we include an in depth discussion of recent work in game-theoretic economic mechanism design for accountable cloud computing.
Ludwig Maximilians Universitat, Munich, Germany
Economic Observations of the Leibniz Supercomputing Centre
The Leibniz Supercomputing Centre (LRZ) is home of SuperMUC-NG, one of the top 10 supercomputers of the world, and offers its computing capabilities to scientists in Bavaria, Germany and Europe. However, its task is more holistic in providing generic IT-services for sciences, including everything from the desktop of the scientist via arbitrary cloud services to the large-scale high performance computer. This talk addresses the economical aspects of LRZ as an example from academia, where key performance indicators are scientific output and not stakeholder revenue. We will take a look at actual costs of the services, accounting of user activities, and sustainability measures for the future. As power usage is a large cost factor, LRZ is also leading in energy-efficient computing with hot-water cooling and heat reuse.
Data Center Growth, Challenges, and Inefficiencies in a Connected World
The rise of smart cities and highly connected mobile devices is driving enormous growth in the data center industry. Data centers are the fundamental infrastructure that supports smart, distributed and connected systems; they currently consume 3% of the global electrical supply - this is expected to double every four years.
Like the smart systems that they support, data centers are highly complex systems-of-systems with interacting hardware, software, power, and thermal components connected to a wide range of service and business models. There is a huge need to address the growing power consumption of the industry (and its resulting financial and environmental impact) but effective solutions are challenging, requiring intelligent and automated reasoning across extremely large volumes of data in a range of disciplines.
This talk focuses on the importance of data centers in supporting modern smart systems, and highlights their growth and inefficiencies. We examine a range of holistic solutions that mark some of the first steps towards managing both the complexity and inefficiencies of data centers. We finish with a case study showing how with advanced behavioural modelling, a scheduling mechanism has been developed that has led to significant power reductions in a data center facility located in northern Sweden.