原理

 写操作先缓存到ssd硬盘上,然后通过一定策略写到普通硬盘上;

 读操作热点数据可以缓存到ssd硬盘上,提高读取数据的速度。

 

软件 ssd 缓存开源软件主要有bcache flashcache

 

bcache 需要编译最新的内核,要求比较高,配置比较复杂,目前主要用于测试环境;

flashcache facebook内部使用的ssd cache软件,开源出来供大家使用,目前有人编译出rhel6rpm包,在centos6上可以直接安装使用。

 

本次初步测试了flashcache

 

测试环境

sdc ssd 硬盘     160G*4 raid 10     fio测试 4k iops 5433

sdb 普通sas硬盘 146G*2 raid  1    fio测试 4k iops 447

 

初步简单测试结果:

    4k 随机写可以达到 5014

    4k随机读可以达到45874

 

flashcache 安装配置方法

 

安装

rpm --import http://elrepo.org/RPM-GPG-KEY-elrepo.org

rpm -Uvh http://elrepo.org/elrepo-release-6-5.el6.elrepo.noarch.rpm

yum install flashcache-utils kmod-flashcache

 

配置:

sdc sdb的缓存

 

flashcache_create -p back  cachedev /dev/sdc /dev/sdb    

cachedev cachedev, ssd_devname /dev/sdc, disk_devname /dev/sdb cache mode WRITE_BACK

block_size 8, md_block_size 8, cache_size 0

Flashcache metadata will use 1192MB of your 24016MB main memory

 

可以用3种方式:

 

 

Writethrough - safest, all writes are cached to ssd but also written to disk

immediately.  If your ssd has slower write performance than your disk (likely

for early generation SSDs purchased in 2008-2010), this may limit your system

write performance.  All disk reads are cached (tunable). 

数据同时写到ssd和普通硬盘

 

Writearound - again, very safe, writes are not written to ssd but directly to

disk.  Disk blocks will only be cached after they are read.  All disk reads

are cached (tunable).

数据同时绕过ssd,直接写到普通硬盘

Writeback - fastest but less safe.  Writes only go to the ssd initially, and

based on various policies are written to disk later.  All disk reads are

cached (tunable). 

数据先写到ssd,随后写到普通硬盘

 

 

 

查看

 flashstat

======================================================================================================

Flashstat: a tool for flashcache status per second

Author   : NinGoo([email protected])

Version  : 0.3

======================================================================================================

          SSD Device:   /dev/sdc         Disk Device:   /dev/sdb          Cache Mode: WRITE_BACK

            Capacity:    303998M          Block Size:         4K     Meta Block Size:      4096b

        Total Blocks:   77823488       Cached Blocks:          9      Cached Percent:          0

         Set Numbers:        512        Dirty Blocks:          0       Dirty Percent:          0

           cache_all:          1      reclaim_policy:       FIFO    dirty_thresh_pct:         20

   max_clean_ios_set:          2 max_clean_ios_total:          4     skip_seq_thresh:         0K

======================================================================================================

          time  read/s write/s diskr/s diskw/s  ssdr/s  ssdw/s uread/s uwrit/s metaw/s clean/s  repl/s wrepl/s   hit%  whit% dwhit%

03-13 14:51:20       0       0       0       0       0       0       0       0       0       0       0       0   0|10    0|0    0|0

 

 

 

fdisk -l /dev/mapper/cachedev  可以看到一个设备

 

Disk /dev/mapper/cachedev: 146.8 GB, 146778685440 bytes

255 heads, 63 sectors/track, 17844 cylinders

Units = cylinders of 16065 * 512 = 8225280 bytes

Sector size (logical/physical): 512 bytes / 512 bytes

I/O size (minimum/optimal): 512 bytes / 512 bytes

Disk identifier: 0x0002da0a

 

               Device Boot      Start         End      Blocks   Id  System

 

使用:

 

可以当一个块设备来使用,划分vg lv

 

 pvcreate /dev/mapper/cachedev

  Physical volume "/dev/mapper/cachedev" successfully created

 vgcreate ssdcachetest /dev/mapper/cachedev  

  Volume group "ssdcachetest" successfully created

 lvcreate -L 100G -n ssdcache_test_lv1 ssdcachetest

  Logical volume "ssdcache_test_lv1" created

 

 

删除

dmsetup  remove cachedev

flashcache_destroy /dev/sdc


欢迎加我微信(ID:xiaolikvm)一起交流。欢迎加入云技术社区KVM/QEMU技术交流QQ,请加QQ:502207183,注明姓名,城市,及要加入KVM群。