RAID模式下性能调优策略

早日实现财富自由-2025

于 2024-10-12 10:17:02 发布

阅读量189

点赞数 7

文章标签： python linux 开发语言 RAID

本文链接：https://blog.csdn.net/lwexin/article/details/142871063

版权

# I/O scheduler
    Set to
     Check: cat /sys/block/sdc/queue/scheduler
     Set: echo noop > /sys/block/sdc/queue/scheduler
# nr_requests
     Check: cat /sys/block/sdc/queue/nr_requests
     Set: echo 512 > /sys/block/sdc/queue/nr_requests
# queue_depth
     Check: cat /sys/block/sdc/device/queue_depth
     Set: echo 512 > /sys/block/sdc/device/queue_depth
# max_sectors_kb
     Check: cat /sys/block/sdc/queue/max_sectors_kb
     Set: echo 512 > /sys/block/sdc/queue/max_sectors_kb
# Linux IO Read-Ahead
     Check: blockdev --getra /dev/sdc
     Set: blockdev --setra 0 /dev/sdc
# rq_affinity
     Check: cat /sys/block/sdc/queue/rq_affinity
     Set: echo "0" > /sys/block/sdc/queue/rq_affinity

Newer kernels (e.g. RHEL 6.4) allow this setting to be '2', which will cause the block layer softirq to run on the exact same CPU that issued the I/O. If running newer kernels then use: echo "2" > /sys/block/sdc/queue/rq_affinity

# rotational
Check: cat /sys/block/sdc/queue/rotational
Set: echo "0" > /sys/block/sdc/queue/rotational