Alexander Yee's writeup is interesting – CPU utilisation was only about 12%, they encountered quite nasty I/O bottlenecks (particularly for writes.) http://www

Alexander Yee's writeup is interesting – CPU utilisation was only about 12%, they encountered quite nasty I/O bottlenecks (particularly for writes.) http://www.numberworld.org/blogs/2019_3_14_pi_record/ http://www.numberworld.org/blogs/2019_3_14_pi_record/ Contradicts the Google blog a little, especially where he points out that they hit performance issues with live migration (Google said it worked fine without impact on the application.)