AutoNUMA: the other approach to NUMA scheduling
Posted Mar 28, 2012 15:04 UTC (Wed) by hamjudo
Parent article: AutoNUMA: the other approach to NUMA scheduling
If it really leads to a 20% performance improvement, on any chip with more than 6 cores, we still come out ahead if we dedicate a whole core in each package to scheduling and tracking memory. Quite a win on the 8 and 12 core packages used in the high end servers.
to post comments)