Modern data processing systems require optimization at massive scale, and using machine learning to optimize these systems (ML-for-systems) has shown promising results. Unfortunately, ML-for-systems is subject to over generalizations that do not capture the large variety of workload patterns, and tend to augment the performance of certain subsets in the workload while regressing performance for others. In this paper, we introduce a performance safeguard system, called PerfGuard , that designs pre-production experiments for deploying ML-for-systems. Instead of searching the entire space of query plans (a well-known, intractable problem), we focus on query plan deltas (a significantly smaller space). PerfGuard formalizes these differences, and correlates plan deltas to important feedback signals, like execution cost. We describe the deep learning architecture and the end-to-end pipeline in PerfGuard that could be used with general relational databases. We show that this architecture improves on baseline models, and that our pipeline identifies key query plan components as major contributors to plan disparity. Offline experimentation shows PerfGuard as a promising approach, with many opportunities for future improvement.

Facial and head movements have important linguistic roles in American Sign Language (ASL) and other sign languages. Without being properly trained, both human observers and existing emotion recognition tools will misinterpret ASL linguistic facial expressions. In this study, we capture over 2,000 photographs of 15 participants: five hearing, five Deaf, and five Children of Deaf Adults (CODAs). We then analyze the performance of six commercial facial expression recognition services on these photographs. Key observations include poor face detection rates for Deaf participants, more accurate emotion recognition for Deaf and CODA participants, and frequent misinterpretation of ASL linguistic markers as negative emotions. This suggests a need to include data from ASL users in the training sets for these technologies.