Shulou(Shulou.com)06/02 Report--

What is vSAN

VSAN,Virtual SAN (Virtual Storage area Network) is a software-defined storage architecture proposed by vmware.

VSAN uses software to turn a large number of blank disks on the server into a virtual shared storage.

VSAN is a clustering function. Through vSAN, the local disks of many hosts are created to create a distributed RAID (depending on different storage policies, it can be RAID1, RAID5/6, etc.), and the data of the virtual machine will be stored on this RAID disk. This disk is actually local to each server, but through the management of vSAN software, it can achieve almost all the functions of virtual shared storage.

The essence of vSAN: integrate the local disks on multiple x86 servers with ESXi system to form a disk group, which is managed by vSphere and vCenter and connected through an internal 10-gigabit network to form a large-capacity storage, which is called vSAN storage (vsanDatastore), and then uses certain policies (storage policies) on this storage, such as using network RAID1 (default policy) to establish and organize virtual machines. Avoid single point of failure to achieve high availability, etc.

The vSAN architecture does not need the external storage of the traditional storage architecture, and the integrated utilization of the storage resources of the server is a part of the HCL (super-converged) architecture.

HCL architecture: integrate the computing, storage and network on multiple servers to avoid the network delay caused by connecting to the external environment, and achieve the effective use of server resources.

In the previous experimental environment, we used eight hard drives on four ESXi servers (each 100G ssd+1T ssd, a total of 400G cache, 4T capacity), through vSAN, integrated into a shared virtual storage with 4T capacity.

II. VSAN requirements

VSAN is integrated into the vsphere cluster environment in an embedded way. As long as certain requirements are met and preparations are made, such as disk preparation and network configuration, vSAN can be enabled through simple settings.

Host requirement

A standard vSAN cluster requires at least 3 ESXi hosts to be configured to FTT=1 (maximum allowable number of failures), but does not guarantee the application of some storage policies, so generally speaking, it is necessary to prepare 4 or 5 ESXi hosts to meet the basic lab requirements. Each server must meet the requirements of the corresponding list, through the list to select the appropriate compatible devices, in addition, each server should have no less than 8 gigabytes of memory.

The hardware of these hosts must be compatible with no less than 8 gigabytes of memory

In our previous experimental environment, we used four ESXi hosts to virtualize using vmware-Workstation, but even so, there was a compatibility warning for the SCSI controller. The CPU of each host is 4 cores and the memory is 8 GB.

Disk requirement

In addition to the disks installed with the esxi system, at least one ssd must be configured as the cache disk on each server, and several other (≥ 1) ssd disks (all-flash architecture) or HDD (hybrid architecture) as data disks. These disks (cache disks + data disks) are referred to by vSAN as a disk group (excluding the disks on which the esxi system is installed).

These disks must be delivered by pass-through or RAID0, that is, physical disks cannot be made into an array of disks other than hardware RAID0 to provide vSAN disks.

However, not every node in the vSAN cluster needs to provide disk groups, and hosts without local storage can still take advantage of vSAN distributed data storage composed of other hosts in the cluster.

These disks must also meet certain conditions before they can use vSAN.

In our previous lab environment, each ESXi host was configured with a 100g ssd as a cache disk (which must be marked as ssd, otherwise it cannot be used as a cache disk) and a 1T ssd (which can also be HDD, but marked ssd to simulate an all-flash environment) as a capacity disk.

VSAN disk configuration

Network requirement

VSAN with all-flash architecture recommends using 10Gb's 10 Gigabit network, and it is not recommended to run vSAN on networks less than 1GB gigabit.

Two 10 Gigabit network cards used to run vSAN

VSAN with mixed architecture can use 1Gb's gigabit network.

III. Basic terminology of vSAN

Object

VSAN is object-based storage, and the object is a separate block device. In the case of not using vSAN, the virtual machine is composed of a group of filegroups, but after using vSAN, the virtual machine is divided into multiple object storage blocks, including: virtual machine home page space, virtual machine exchange files, vmdk (Hard disk) and so on. In our experimental environment, our newly established centos8 virtual machine includes two objects, the virtual machine home directory and Hard disk 1 (if there are two disks, there are three objects).

Object

Add a new hard drive and you will have three objects.

3 object

module

VSAN refers to what data each object has. VSAN makes a network RAID 1 (under the default storage policy) for each object of the virtual machine, that is, it mirrors the virtual machine home directory, vmdk, and so on, and has two components. This allows them to have copies on multiple hosts for the purpose of high availability. These images are called component in vSAN, and the number of components is related to the storage policy.

The following figure shows the three components included in Hard disk 1:

3 components of Hard disk 1

You can see that it contains three components (witness is also a component, which will be described in more detail later), of which component 1 is placed on host 32, component 2 is placed on host 33, and witness is placed on 34.

Component 1 and component 2 store metadata, they are mirrored, and the content is exactly the same.

Similarly, Hard disk 2 and the virtual machine home page each have three components, which are also stored on different hosts.

Three components of the virtual machine home page

Copy

It is the number of components (excluding witnesses) that vSAN uses RAID to achieve high availability, so there are multiple copies of an object to avoid a single point of failure. Similarly, the number of copies is related to the storage policy. We used RAID1 in the experimental environment above, with 2 copies.

Number of copies

Witness (Witness)

Generally understood as "arbitration", it is a special component of RAID1 (RAID5/6 does not witness this component). It does not include metadata and is only used to determine how to recover when arbitration occurs after a vSAN failure. In some cases, this witness can be placed on a specific host outside the cluster (such as in a two-node cluster or an extended cluster), which is called the witness host.

Disk group

After the cache disk and capacity disk on each host are declared, they form a disk group, in which the cache disk must be ssd disk, the capacity disk can be ssd, SAS or ordinary disk, the capacity disk is ssd or all-flash architecture, and the capacity disk is not ssd but hybrid architecture.

In our experimental environment, each host consists of a 100g cache disk and a 1T capacity disk, which is an all-flash architecture (note that only vSAN Enterprise Edition supports all-flash architecture, not the standard version).

Disk groups, cache disks, and capacity disks

Storage strategy

VSAN is storage policy-based management (Storage Policy-Based Management) that indicates how virtual machines are stored in vSAN, such as how many host failures are allowed, whether RAID1 or RAID5/6 is used, how many stripes (data segmentation), and so on.

The core of the management vSAN based on the storage policy, all virtual machines deployed on the vSAN must use a storage policy. "if no new storage policy is created, the virtual machine uses the default policy." A special article will be devoted to vSAN's storage strategy later.

The new virtual machine centos8 in our experimental environment uses the default storage policy. Let's take a look at its details.

Default storage policy

As you can see, under the default storage policy:

Number of failures allowed (FTT): 1 (this is why at least 3 ESXI hosts are required)

Storage method: RAID 1 (mirroring), mirroring the objects of each virtual machine, with 2 copies (each copy includes 2 components and 1 witness)

Number of stripes: 1 (no stripes are used)

Using streamlined provisioning

Number of stripes on object disk (Number of disk stripes per object)

The virtual machine object is divided into multiple stripes (stripe) and stored on multiple hard disks, which helps to improve the performance of data reading. In the example shown below, the virtual machine shown is divided into two stripes an and b for storage; because Raid-1 is also used, the virtual machine object is saved with two copies, one in the disk group of the server ESX01 and the other in the disk group on the server ESX02 and ESX03.

The number of stripes of an object is actually the concept of Raid-0, but when we define the configuration of Raid-0, we use the concept of stripe width (stripe width). Stripe width refers to the size of each stripe block, and the number of stripes depends on the number of disks that make up the Raid-0 array. The default number of object stripes in vSAN is 1 (it is recommended that you do not change this value), with a maximum of 12. Setting this value to greater than 1 can improve system performance, but will cause more resource consumption.

In our experimental environment, we create a new storage policy, using the policy of RAID1, stripe number = 2

Create a new virtual machine and apply this strategy

Placement information of virtual machine

As you can see, Hard disk 1 actually makes a RAID 01 (note that it is not RAID 10)

Add a picture of RAID 01

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.