Extensive benchmarking tests on DECTRIS CLOUD, conducted following James Holton’s protocols, demonstrate how the platform accommodates various computational needs for macromolecular crystallography (MX) data analysis. These results highlight the performance range across DECTRIS CLOUD’s instance offerings, with the Lightspeed instance achieving top placement among XDS benchmarks.
The DECTRIS CLOUD team has successfully conducted benchmarking tests for the macromolecular crystallography (MX) program XDS. The tests followed James Holton’s widely recognized protocols, offering detailed comparisons of processing times for XDS workflows across DECTRIS CLOUD’s instance types.
The benchmarks involved processing a simulated 3600-image dataset of lysozyme from a Pilatus-sized detector. The dataset ensures consistency across benchmarking efforts and provides a meaningful comparison of processing times. We evaluated the speed of processing using all key stages: INIT, COLSPOT, IDXREF, INTEGRATE, and CORRECT. Optimal settings for parameters such as MAXIMUM_NUMBER_OF_JOBS, MAXIMUM_NUMBER_OF_PROCESSORS, and NUMBER_OF_IMAGES_IN_CACHE were applied at each stage to ensure peak performance. The results placed the Lightspeed configuration (AMD EPYC 9R14, 192 CPUs, 768 GB RAM) at the top of the James Holton XDS benchmark list with a total time of 72s, demonstrating the platform’s capacity for high-performance data analysis.
The accompanying graph depicts the relationship between processing time and the number of processors used for XDS, providing insight into the performance of DECTRIS CLOUD. Additionally, the graph includes data from the 10 best-performing benchmarks on James Holton’s list for comparison. Based on the fits log(Time)= m log(#CPUs) + log(C) , both datasets exhibit similar scaling behavior with increasing processing power, as indicated by their scaling exponents of approximately m ~-0.5. However, the DECTRIS CLOUD machines demonstrate a base performance parameter, log(C), that is ~10% lower than the other dataset, suggesting a performance advantage at lower processor counts.
By providing a transparent performance evaluation, DECTRIS CLOUD supports MX facilities as they adopt cloud-based solutions for increasingly complex structural biology studies.
For additional details on how DECTRIS CLOUD can enhance your lab’s workflow, contact us!
