Thank you for reporting this. I'm currently evaluating this with some benchmarking experiments and my current observation concurs with the observations. I believe enabled=madvise is probably the best way forward but I still have several more test scenarios to complete to ensure we don't hit any unexpected performance regresions.
Thank you for reporting this. I'm currently evaluating this with some benchmarking experiments and my current observation concurs with the observations. I believe enabled=madvise is probably the best way forward but I still have several more test scenarios to complete to ensure we don't hit any unexpected performance regresions.