Shulou(Shulou.com)06/03 Report--

VSAN uses Storage Policy-Based Management (SPBM, Chinese translation for "storage policy-based management") to deploy virtual machines.

By using storage policy-based management, virtual machines can apply different policies according to the needs of the production environment and without shutdown.

All virtual machines deployed on vSAN must use a storage policy, and if no new storage policy is created, the virtual machine will use the default policy.

There are the following main types of vSAN storage policies:

1. Number of Failures to Tolerate (allowable number of failures)

Number of Failures to Tolerate, abbreviated as FTT, is translated as "allowable number of failures" in Chinese.

This policy defines the maximum number of failures allowed in the event of a failure of a host, disk, or network in the cluster without affecting the operation of the virtual machine, and the FTT value is 1 by default.

The value of FTT determines the number of components (replicas) and the number of ESXi hosts required in the virtual machine storage objects in the vSAN cluster. Assuming that the value of FTT is set to n, there will be 1 copy of FTT, which requires at least one 2n+1 host. If under the default policy, FTT=1, there will be 2 copies and 3 hosts will be required. The details are as follows:

Table: FTT values correspond to the number of copies, witnesses, and ESXi hosts

FTT copy witnessed the number of ESXi hosts 0101121323253437

Our experimental environment is 4 ESXi hosts, which can not meet the number of hosts required by FTT=2, so we can only FTT=1.

II. Number of Disk Stripes per Object

Number of Disk Stripes per Object, abbreviated as Stripes, translates to "disk stripes per object", which means that the components of the Hard disk (the virtual machine home page does not use stripes) of the storage object are divided into several parts and distributed on multiple hosts to improve its read and write performance.

As shown in the figure below, a virtual machine uses Hard disk 1 with number of stripes = 2

Component 1 is divided into two stripes, these two stripes are placed on different hosts, and they form component 1 together in the form of RAID0. When reading and writing data, it can read and write from two stripes in parallel, thus improving the efficiency.

If you set stripe = 3, the above figure adds a "component 1 stripe 3", and component 1 is divided into 3 stripes.

Similarly, component 2 is divided into two stripes.

The contents of component 1 and component 2 are exactly the same, mirroring each other and forming the RAID1.

The Stripes value is equivalent to the environment of RAID0 and is distributed across multiple physical disks. Generally speaking, the default value of stripes is 1, and the maximum value is 12.

If the parameter value is set to greater than 1, the virtual machine can get better IOPS performance, but will consume more system resources. The default value of 1 is sufficient for most virtual machine loads.

For disk I / O intensive operations, the Stripes value can be adjusted. When the size of an object exceeds 255GB, the system forcibly splits the object even if Stripes is 1 by default.

III. Flash Read Cache Reservation

Flash Read Cache Reservation, translated as "flash read cache reservation" in Chinese. The default is 0.

This parameter is combined with the disk size of the virtual machine to set the Read Cache size, which can be calculated as a percentage, which can be accurate to 4 decimal places. If the disk size of the virtual machine is 100GB and the flash read cache reservation is set to 10%, the flash read cache reservation value will use the SSD capacity of 10GB. When the virtual machine disk is larger, it will take up a lot of flash space.

In a production environment, flash read cache reservation is generally not configured because the flash read cache reserved for virtual machines cannot be used for other objects, and unreserved flash memory can be shared with all objects.

It should be noted that Read Cache fails in an all-flash environment.

IV. Force Provisioning

It is translated as "compulsory provision" in Chinese.

When mandatory provisioning is enabled, vSAN monitors the storage policy application. When the storage policy cannot be satisfied (such as enabling FTT=2 under the condition of 3 hosts), if mandatory provisioning is selected, the policy will be ignored and forcibly set to:

FTT=0

Stripe=1

Object Space Reservation=0

This situation should be avoided because no failures are allowed in the entire vSAN environment.

5. Object Space Reservation

Object Space Reservation, abbreviated as OSR, is translated as "object space reservation" in Chinese.

The default is 0, which means that the disk mode of the virtual machine is Thin Provisioning, which means that no space will be reserved when the virtual machine is deployed, and the space will be used only when the virtual machine storage grows.

If OSR is set to 100%, the capacity requirement for virtual machine storage will be reserved in advance, that is, Thick Provisioning (thick setting).

It should be noted that for Thick Provisioning in Virtual SAN, there is only Lazy Zeroed Thick (thick setting delay zeroing, LZT), but no Eager Zeroed Thick (thick setting zeroing, EZT).

VI. Fault tolerance

Fault tolerance is a new virtual machine storage strategy introduced from vSAN version 6.2. it is mainly to solve the problem that the old version of vSAN uses RAID 1 technology to take up a lot of disk space (mainly the introduction of RAID5/6).

VSAN version 6.7 continues to be optimized to provide more vSAN storage space.

VII. Object IOPS restrictions

Object IOPS restriction is a virtual machine storage policy that has been improved since vSAN version 6.2, which can impose different IOPS restrictions on virtual machines according to application requirements, and improve the efficiency of IOPS.

Eight. disable object checksum

The purpose of disabling object checksum is to ensure the integrity of vSAN data. The system will check the check data during read and write operations, and repair the data if there is something wrong with the data.

Disable the object checksum setting to NO, and the system will fix the problem data; if set to YES, the system will not fix the problem data.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.