Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn systematic approaches to optimizing sequential tasks in computer systems through this 15-minute conference talk from OSDI '25. Discover three fundamental principles for performance optimization—task removal, replacement, and reordering—and explore how they translate into eight actionable methodologies including batching, caching, precomputing, deferring, relaxation, contextualization, hardware specialization, and layering. Examine how researchers from Georgia Institute of Technology analyzed a decade of OSDI and SOSP papers to validate these optimization strategies comprehensively account for observed sequential performance improvements. Explore practical applications through two detailed case studies focusing on file and storage systems optimization and kernel synchronization analysis to identify missed optimization opportunities. Gain insights into SysGPT, a specialized fine-tuned GPT model trained on curated systems literature that provides context-aware performance suggestions with greater specificity and feasibility compared to general-purpose language models, demonstrating its potential as an optimization assistant for systems researchers and practitioners.
