Dell VxRail Architecture Overview

Sizing considerations

While sizing the VxRail stretched cluster with VMware vSAN, calculate host and witness requirements, fault domains (FD), and witness use.

Host requirement calculation

The storage policies applied to a stretched cluster define the minimum number of hosts that are required per site. For example, a site disaster tolerance set to dual site mirroring (stretched cluster) and a failure to tolerate set to 1 failure - RAID - 5 (erasure coding), results in three data blocks and one parity component. At least four hosts per site are required. The final configuration would be 4+4+1, four hosts per site, and one witness host.

From a capacity standpoint, if you have a 100 GB VM and set the site disaster tolerance to dual site mirroring (stretched cluster) and a failure to tolerate equal to 1 failure - RAID - 1 (mirroring) that means RAID 1 is set in each site. A 100 GB VM requires 200 GB in each location which means 200 percent required local capacity and 400 percent for the total cluster. RAID 5 and RAID 6 are only available when using all-flash. Using the VMware vSAN ESA RAID 5 storage policy with six defined FDs, every 100 GB of stored data consumes 125 GB of raw capacity. That is a 25 percent savings in capacity over the similar cluster running VMware vSAN OSA with better performance.

The following table provides the capacity per site and the overhead:

Table 1. Capacity per siteCapacity per site
VMware architecture	FTT	RAID	Placement	Capacity consumption	Recommended minimum hosts
VMware vSAN OSA and VMware vSAN ESA	1	1	3x Mirror	2x	4
VMware vSAN OSA	1	5	3+1	1.33x	5
VMware vSAN ESA	1	5	4+1	1.25x	6
VMware vSAN ESA	1	5	2+1	1.5x	4
VMware vSAN OSA and VMware vSAN ESA	2	1	3x Mirror	3x	6
VMware vSAN OSA	2	6	4+2	1.5x	7
VMware vSAN ESA	2	6	4+2	1.5x	7

The two RAID 5 erasure coding schemes in VMware vSAN ESA are:

RAID 5 works on a 3-node cluster, four hosts are recommended. There is a performance advantage to RAID 5 over RAID 1 with the capacity consumption offering a better return.
VMware vSAN ESA RAID 5 automatically adapts to the cluster size, with its new adaptive RAID-5 erasure code to best suit the host count in the cluster. The adaptive feature adjusts the RAID 5 scheme based on the host count of the cluster.

For larger clusters, use RAID 6. For VMware vSAN ESA, do not use RAID 1 unless the cluster is 2-node.

Cluster compute resource use

For full availability, Broadcom recommends running at 50 percent of resource consumption across the VMware vSAN stretched cluster. If a complete site failure occurs, all the VMs can run on the surviving site.

Broadcom understands that you may want to run levels of resource utilization higher than 50 percent. While you can run at higher utilization in each site, in the event of failure, not all VMs are restarted on the surviving site.

Table 2. Stretched cluster capacitystretched cluster capacity
Protection	Architecture	PFTT	FTM	SFTT	Percentage of VM
Protection	Architecture	PFTT	FTM	SFTT	Preferred site capacity	Secondary site capacity	Total capacity
Across Sites Only	VMware vSAN OSA	1	Mirroring	0	100	100	200
Mirroring (RAID 1 Single Failure)	VMware vSAN OSA	1	Mirroring	1	200	200	400
Mirroring (RAID 1 Double Failure)	VMware vSAN OSA	1	Mirroring	2	300	300	600
Mirroring (RAID 1 Triple Failure)	VMware vSAN OSA and VMware vSAN ESA	1	Mirroring	3	400	400	800
Erasure Coding (RAID5/ Single Failure)	VMware vSAN OSA	1	Erasure coding	1	133	133	266
Erasure Coding(RAID5/ Single Failure)	VMware vSAN ESA	1	Erasure coding	1	125	125	250
Erasure Coding (RAID6/ Double Failure)	VMware vSAN OSA	1	Erasure coding	2	150	150	300
Erasure Coding (RAID6/ Double Failure)	VMware vSAN ESA	1	Erasure coding	2	150	150	300

Fault domains

FDs provide core functionality of the VMware vSAN stretched cluster. The supported number of FDs in a VMware vSAN stretched cluster is three. The first FD can be addressed as the preferred data site. The second FD can be addressed as the secondary data site, and the third FD is the witness host site. If the preferred or secondary site goes offline, keep data site use below 50 percent to ensure proper availability.

Witness appliance

The VMware vSAN witness appliance provides the following features:

Deployed on a VMware ESXi host. The VMware vSAN witness appliance includes licensing, while a physical host must be licensed accordingly.
Provided with each release of VMware vSAN. The underlying VMware vSphere version is the same as the version running VMware vSAN. During deployment of the appliance, verify that you have the same versions of VMware vSAN witness appliance and VMware vSAN.
Deploying the witness for a normal configuration requires 10 GB of cache device.
The physical host does not require a flash nor an SSD device.
The device capacity must be 350 GB or greater. Traditional spinning drives are sufficient because the witness marks those devices as required.

Table 3. Witness sizingwitness appliance sizing
Witness	VM	Boot/cache/capacity	Witness components
Tiny	10	12/10/15 GB	750
Normal	<500	12/10/350 GB	2100
Large	>500	12/10/700 GB	4500

Use the calculator in Appendix B to determine if the size of the appliance is sufficient.

Each VMware vSAN stretched cluster configuration requires a witness host. The witness must reside on a third site that has independent paths to each data site. While the witness host must be part of the same VMware vCenter Server as the hosts in the data sites, it must not be on the same cluster as the data site hosts. Network Address Translation (NAT) is not supported.

NOTE:The witness host that is deployed as a VM using the VMware vSAN witness appliance has an embedded license that is automatically applied during deployment.

Witness host requirements

The following witness host requirements apply:

Verify that the witness host is not part of any VMware vSAN enabled cluster.
Enable one VMkernel adapter with VMware vSAN traffic and connectivity for the witness host.
Connect the adapter to all hosts in the stretched cluster.

Welcome

Welcome to Dell