This section provides an architecture overview and guidance on managing and scaling a VMware Horizon 7 environment on Dell EMC VxBlock System 1000.
The following figure depicts the architecture of a representative, recommended VDI design based on VxBlock 1000, including the network, compute, management, and storage layers. This architecture aligns with the VMware Horizon pod and block design. A pod is divided into multiple blocks. Each block has one or more VMware vSphere clusters, a VMware vCenter Server, and, for linked clones, a Composer server.
Deployment of this Ready Architecture supports all VMware cloning techniques, including full, linked, and instant cloning.
A vSphere cluster can have a maximum of 64 nodes and 8,000 virtual machines (VMs) per cluster. To expand beyond these limits, you can add clusters and balance the VMs and nodes across the new clusters.
For more information about Horizon pods and blocks, see Horizon 7 Enterprise Edition Reference Architecture.
Solutions based on VxBlock 1000 provide flexibility as you scale, reducing the initial and future cost of ownership. You can expand VxBlock 1000 by adding compute, network, and storage resources. After the initial deployment, you can add more disks and arrays for increased capacity and more servers for increased performance. Add physical and virtual servers to the server pools to scale horizontally. Add resources to the infrastructure, for example, CPU, memory, or SSD drives, to scale vertically.
Each component of the solution architecture scales independently, depending on the required number of supported users. VxBlock Advanced Management Platform (AMP) and compute domains can both be scaled out or expanded independently to accommodate their respective workloads. You can add VxBlock 1000 management, compute, networking, and storage components at any time to expand capabilities in a granular fashion.
The AMP-3S environment is scalable from 2 to 16 compute nodes while the compute environment can be scaled from 3 to 616 half-width B-Series servers or 3 to 1,120 C-Series servers with Cisco UCS fourth generation network fabrics.
When scaling the AMP-3S environment, consider both the VxBlock management components and the VDI management components. The following table outlines the AMP scaling guidance using the following assumptions:
Number of VDI users
Number of Cisco UCS C220 M5 servers for AMP-3S environment
The boundary for a Horizon block is the vCenter. The number of VMs that a vCenter (and, therefore, a block) can host depends on the type of Horizon 7 VMs being used. The recommended limits for a Horizon block are as follows:
This Ready Architecture for VDI uses instant clones, as shown in the following figures.
VMware recommends a limit of 5,000 instant-clone VMs per block. With these limits, 50 compute nodes with 100 knowledge user VMs per node is the maximum number of VMs for the block.
The following figure shows a 5,000-user Horizon block based on a 100-user-per-node density.
The following figure shows a scale-out to a 20,000-user Horizon pod with 5,000 user blocks. The AMP-3S HA cluster contains the vCenter Server instances and VDI management components.
Multiple aspects of the VxBlock 1000 system can be scaled up, including the compute layer's CPU, memory, and I/O; the storage layer's SSDs; and the network layer's fabric resources. When designing a VxBlock 1000 system, consider the number of users that the system will eventually support to ensure the proper selection of components such as the Cisco MDS switches. In this example, selection of a switch with higher port density provides for nondisruptive expansion for the environment in the future.
The Dell EMC XtremIO X2 array can be scaled up by adding as few as six SSDs at a time to a single X-Brick block.
The AMP-3S configuration cannot be scaled up within the context of CPU, memory, or I/O.
Note: The VxBlock 1000 compute layer configurations can have additional or faster processors or memory than the guidance we provide here. Maximum scalability is determined by the architecture of the selected compute model.
For more information about Horizon pod and block architecture, and scaling, see the VMware Workspace ONE and VMware Horizon 7 Reference Architecture.
Advanced Management Platform (AMP)
The AMP is an integral part of VxBlock 1000. It provides a dedicated location at which to install and support the management software that helps to install and operate components within the VxBlock 1000 system. The AMP consists of a dedicated set of hardware and software that is set up and managed independently of the computing domains within the system. The hardware has a fixed CPU and memory configuration. This guide discusses AMP-3S in detail, AMP-VX is also an option.
AMP-3S consists of the following high-level hardware components:
AMP-3S consists of the following software components:
The AMP infrastructure is designed to be fault tolerant in terms of server, storage, and network design. The network is designed for fault isolation from the VxBlock 1000 workload domain.
The AMP infrastructure is also the appropriate location to host the VDI management components, including:
For more information about these components, see the VMware Horizon section.
This design guide describes a single-site or single-data-center design. For multisite or disaster recovery (DR) configurations, see the VMware Workspace ONE and VMware Horizon Reference Architecture.
In typical enterprise computing environments, VDI is not the only workload used. Many other workloads and related applications support different business units within an organization. The VxBlock infrastructure with all-flash storage provides IT organizations with a single, complete platform to simultaneously support mixed workloads and mixed applications, without requiring you to modify the application itself or use proprietary application tools to unlock performance.
IT organizations no longer need to separate workloads with different I/O patterns or different uses (such as production and test) on separate infrastructure silos. VxBlock 1000 can consolidate workloads with mixed I/O patterns onto a single infrastructure. The reduced cost of configuration, support, and maintenance lowers the overall TCO of running an enterprise data center.
Consider the following information when designing a mixed-workload environment within a single VxBlock system:
Separate VDI workloads from other workloads at the compute layer. In a VDI environment, factors within the compute servers such as context switching, oversubscription, and latency can drastically affect the user experience.
Additional recommendations include:
Logically separate VDI workloads from other workloads at the network layer by using either traditional VLANs and subnets or, in the case of software-defined networks, VXLANs. However, the underlying physical networking fabric can be shared.
Additional recommendations include:
Ideally, separate each workload onto its own correlated physical storage arrays. Select the appropriate Cisco MDS switches to accommodate the maximum ports that are needed for the storage configuration.
Additional recommendations include:
Although not recommended, running all workloads on the same set of storage arrays is possible. When planning and designing the storage layer in such a manner, consider the following:
The AMP-3S management environment is where the VDI workload hosts its own management components. It is also a good location for management components for other workloads that are not compute-intensive.
When you design the AMP, ensure that the management components that support the mixed workloads do not exceed the AMP's maximum processing or storage capabilities. If they do, the management components might have to be designed to run on the VxBlock 1000 computing environment.