Better than a petaflop: the power of efficient experimental design

Abstract

Recent advances in high-performance computing have pushed computational capabilities to a petaflop (a thousand trillion operations per second) in a single computing cluster. This breakthrough has been hailed as a way to fundamentally change science and engineering by letting people perform experiments that were previously beyond reach. But for those interested in exploring the I/O behavior of their simulation model, efficient experimental design has a much higher payoff at a much lower cost. A well-designed experiment allows the analyst to examine many more factors than would otherwise be possible, while providing insights that cannot be gleaned from trial-and-error approaches or by sampling factors one at a time. We present the basic concepts of experimental design, the types of goals it can address, and why it is such an important and useful tool for simulation. Ideally, this tutorial will entice you to use experimental designs in your upcoming simulation studies.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 32 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.