2018 Cheriton Research Symposium

Daniel G. Brown, Acting Director, David R. Cheriton School of Computer Science • DC 1302

Welcome and Opening Remarks

Refreshments will be served

Bernard Wong, University of Waterloo • DC 1302

Consensus Protocols for Blockchains, Transactions, and Cloud Computing

Achieving consensus among a set of distributed entities is a fundamental problem at the heart of many distributed systems. Although there has been a significant amount of past work on building efficient general-purpose consensus protocols, most of these protocols still have performance and cost-related problems that make them unsuitable for many current and emerging applications. In this talk, I will describe our ongoing work on building application-specific consensus protocols. The goal of these protocols is to solve the most critical consensus-related problem for a specific class of applications by exploiting properties common to them or to the environment in which they are deployed, and by trading off features and performance attributes that are of minimal value to these applications.

Our work targets three main classes of applications: Blockchains, geo-distributed transaction processing systems, and fault-tolerant cloud applications. In a blockchain, a large number of participants share a replicated log or ledger. A consensus protocol for blockchains must therefore scale to hundreds of nodes spanning multiple datacenters. For geo-distributed transaction processing systems, consensus must be achieved on the critical path of each transaction to safely replicate the transaction’s state. A consensus protocol for transaction processing systems must provide low worst-case latency in the absence of failures. 

Finally, cost is often the main concern for the many fault-tolerant applications that are deployed in the cloud. A consensus protocol for the cloud can be more cost efficient than a general-purpose consensus protocol by sharing resources between separate consensus groups operating in the same cloud datacenter. I will present our consensus protocols for these classes of applications, and show their effectiveness at meeting their application-specific performance goals.

Edward Lank, University of Waterloo • DC 1302

WRiST: Wearables for Rich, Subtle, and Transient Interactions in Ubiquitous Environments

Modern personal computing devices — tablets, smartphones, smartwatches, fitness trackers, remote controls — represent an ever-more-pervasive component of everyday life. Alongside this trend toward an increase in computing that you bring with you, we also continually encounter devices that exist both in tangible spaces and in the electronic domain. While this computationally augmented reality exists around us, interaction with this embedded, encountered computation rarely achieves the fluidity that one would desire. As a simple example, consider connecting your laptop or smartphone to a data projector. It almost always works well … except when it does not.

Motivated by our earlier work on motion gestures for smartphone input, I will describe on-going research projects leveraging commodity smartwatches as proxies to support WIMP-style input, text input, and rich gestural input to external computation. Because this work lies in the domain of free-space gestural input, I will also discuss our work on enhancing both the reliability and the perceived reliability of gesture recognition for end-users. I will conclude by outlining our emerging vision of wearables and personal devices as a platform for the rich, subtle, and transient interactions necessary in an ever-more ubicomp-based reality.

David R. Cheriton, Stanford University • DC 1302

Rethinking Virtual Memory for a Multi-core Terabyte Memory Computing World   

Virtual memory is now over 50 years old, and has changed relatively little in that time, yet the sizes of DRAM, performance CPUs, and sizes of application datasets have changed by orders of magnitude during that time. Fault-tolerance is now a requirement in many applications. This talk discusses a project at Stanford to rethink virtual memory, including some preliminary results.

Poster Session — David R. Cheriton ​Graduate Scholarship recipients

DC Atrium

Awards Ceremony

DC Fishbowl, Room 1301

Gelato will be served — everyone welcome!