Cloud-based science labs allow science classrooms to engage in authentic science inquiry without the logistical and procedural complexities of real hands-on laboratories. A major challenge is to understand the variation of teaching contexts and needs and the implication for technology design. Utilizing an Biology Cloud Lab (BCL) for real-time experimentation that had been deployed as Massive Open Online Course (MOOC) previously, we now execute three in-depth case studies of how K-12 teachers (with ~200 students) adapt and deploy this BCL and MOOC material for their own class-room needs. We uncovered that teachers differed in their diverse range of learning objectives (e.g., large-scale data analysis vs. computational modeling based on real data), how the teachers were able to adapt the BCL and MOOC material differently to meet their respective curricular needs, and how logistically complex and time intensive processes can be automated to allow teachers and students to focus on key practices. This work reveals the need and design rules for well-standardized and modularized cloud-based science laboratories and accompanying learning materials to then enable rich inquiry based learning activities that work at scale while also being highly customizable to the local classroom contexts.

Big data, specifically data analytics, is responsible for driving many of consumers' most common online activities, including shopping, web searches, and interactions on social media. In this paper, we present the first (micro)architectural investigation of a new industry-standard, open source benchmark suite directed at big data analytics applications—TPCx-BB (BigBench). Where previous work has usually studied benchmarks which oversimplify big data analytics, our study of BigBench reveals that there is immense diversity among applications, owing to their varied data types, computational paradigms, and analyses. In our analysis, we also make an important discovery generally restricting processor performance in big data. Contrary to conventional wisdom that big data applications lend themselves naturally to parallelism, we discover that they lack sufficient thread-level parallelism (TLP) to fully utilize all cores. In other words, they are constrained by Amdahl's law. While TLP may be limited by various factors, ultimately we find that single-thread performance is as relevant in scale-out workloads as it is in more classical applications. To this end we present core packing: a software and hardware solution that could provide as much as 20% execution speedup for some big data analytics applications.