VMware vSAN 6.7 actual combat (c) vSAN storage policy overview

vSAN use Storage Policy-Based Management (referred to SPBM, the Chinese translation for "policy-based storage management") to deploy virtual machines.

By using policy-based storage management, virtual machines can produce according to the needs of the environment and without turning off the different policies.

All vSAN deployed on virtual machines must use a storage policy, if not create a new storage strategy, the virtual machine will use the default strategy.

vSAN storage strategy are the following types:

A, Number of Failures to Tolerate (number of failures allowed)

Number of Failures to Tolerate, referred to as the FTT, the Chinese translation for "number of failures allowed."

The policy definition without affecting the virtual machines running, the cluster hosts, disk or network if a failure occurs, the maximum number of allowed failed, FTT default value of 1.

FTT value determines the number (number of copies) stored in the virtual machine objects vSAN concentrated population and the number of components required ESXi host, the assumed value set FTT is n, then there will be n + 1 copies, requires at least 2n + 1 hosts. As in the default policy, FTT = 1, then there will be two copies, need three hosts. In the following table:

 Table: FTT values ​​corresponding replica, the number of hosts and ESXi witness

FTT Duplicate witness ESXi hosts
0 1 0 1
1 2 1 3
2 3 2 5
3 4 3 7

Our test environment is a four ESXi host, the host can not reach FTT = Number of 52 requirements, we can only FTT = 1.

二、Number of Disk Stripes per Object

Number of Disk Stripes per Object, referred to as the Stripes, the Chinese translation for "disk with the number of each object," represents a storage object Hard disk (virtual machine will not be used strip Home) components are divided into several parts, distribution a plurality of hosts, increase read and write performance.

Below, using a virtual machine is the number of stripes = Hard disk 2 in the case where a

Wherein when the assembly 1 is divided into two strips, the two strips are placed on different hosts, the form of which is composed of RAID0 assembly 1, to read and write data in parallel from the two stripes with read and write, thereby increasing the efficiency.

如果设置条带=3,则上图会增加一个“组件1条带3”,组件1被划分成3个条带。

同理,组件2也被划分成2个条带。

组件1和组件2的内容完全一样,互为镜像,组成了RAID1。

Stripes值相当于RAID0的环境,分布在多个物理磁盘上。一般来说,Stripes默认值为1,最大值为12。

如果将该参数值设置为大于1时,虚拟机可以获取更好的IOPS性能,但会占用更多的系统资源。默认值1可以满足大多数虚拟机负载使用。

对于磁盘I/O密集型运算可以调整Stripes值。当一个对象大小超过255GB时,即使Stripes默认为1,系统还是会对对象进行强行分割。

三、Flash Read Cache Reservation

Flash Read Cache Reservation,中文翻译为“闪存读取缓存预留”。默认为0。

这个参数结合虚拟机磁盘大小来设定Read Cache大小,计算方式为百分比,可以精确到小数点后4位,如果虚拟机磁盘大小为100GB,闪存读取缓存预留设置为10%,闪存读取缓存预留值会使用10GB的SSD容量,当虚拟机磁盘越大的时候,会占用大量的闪存空间。

在生产环境中,一般不配置闪存读取缓存预留,因为为虚拟机预留的闪存读取缓存不能用于其他对象,而未预留的闪存可以共享给所有对象使用。

需要注意的是,Read Cache在全闪存环境下失效。

四、Force Provisioning

中文翻译为“强制置备”。

启用强制置备后,vSAN会监控存储策略应用,在存储策略无法被满足时(如在3台主机的条件下启用FTT=2),这时如果选择了强制置备,则该策略将被忽略,并且强行设置为:

FTT=0

Stripe=1

Object Space Reservation=0

这种情况是要避免的,因为整个vSAN环境已经不允许出现任何故障。

五、Object Space Reservation

Object Space Reservation,简称为OSR,中文翻译为“对象空间预留”。

默认为0,也就是说虚拟机的磁盘模式为Thin Provisioning(精简置备),意味着虚拟机部署的时候不会预留任何空间,只有当虚拟机存储增长时空间才会被使用。

OSR如果设置为100%,虚拟机存储对容量的要求会被预先保留,也就是Thick Provisioning(厚置备)。

需要注意的是,Virtual SAN中Thick Provisioning,只存在Lazy Zeroed Thick(厚置备延迟置零,LZT),不存在Eager Zeroed Thick(厚置备置零,EZT)。

六、容错

容错是从vSAN 6.2版本开始引入的新的虚拟机存储策略,其主要是为了解决老版本vSAN使用RAID 1技术占用大量的磁盘空间问题(主要是引入了RAID5/6)。

vSAN 6.7版本继续进行了优化,提供更多的vSAN存储空间。

七.对象IOPS限制

对象IOPS限制是从vSAN 6.2版本开始完善的虚拟机存储策略,可以对虚拟机按应用需求进行不同的IOPS限制,提高I/O效率。

八.禁用对象校验和

禁用对象校验和是为了保证vSAN数据的完整性,系统在读写操作时会检查检验数据,如果数据有问题,则会对数据进行修复操作。

禁用对象校验和设置为NO,系统会对问题数据进行修复;设置为YES,系统不会对问题数据进行修复。


Guess you like

Origin blog.51cto.com/11694088/2460323