Deployment Planning

HDM Deployment

HDM provides a very flexible deployment model providing options to cater to different user needs in terms of use case, performance, scalability and security. This needs to be selected as part of the installation of on-premises components.

alt_text

HDM Deployment Types

While deploying HDM, users need to select a combination of Deployment Mode and Resource Allocation type. The features of the available options are provided below to guide the user to choose the deployment type that is best suited for their requirements.

Deployment Mode : Standalone or Cluster

Based on scale and failure tolerance requirements, users can choose to deploy either in standalone mode or as a cluster. Please note the following:

  1. In standalone mode, a fixed number of HDM components are installed, regardless of cluster size. In addition, this mode does not include any redundancy in the HDM components, so it will be less fault tolerant.
  2. In cluster mode, some HDM components will be installed based on the scale of the cluster. The delivers greater redundancy, superior fault tolerance, and more concurrent migrations than the standalone option, but it will consume more resources.

Concurrency and Scale Support

The following table highlights the amount of concurrency and scalability supported by each HDM deployment type:

Note: Here, the N in Cluster (N) refers to the number of nodes in the cloud cluster where HDM is deployed.

Concurrent

Cold Migrations

Compute

Migrated VMs

On-Cloud

Concurrent

Warm Migrations

Lite
Standalone 8 10 2
Cluster (N) 8 2 x 10 N x 2
Standard
Standalone 8 20 4
Cluster (N) 8 N x 20 N x 4
Performance
Standalone 8 25 8
Cluster (N) 8 N x 25 N x 8

Note : In Ultra Lite mode, HDM can process multiple simultaneous migration requests. The requests will be analyzed and queued by the system. Eight VMDKs per node can be processed simultaneously.

Resource Allocation : Ultra Lite, Lite, Standard or Performance

HDM provides Ultra LIte, Lite, Standard and Performance deployment options. Each option involves a trade-off between CPU and memory resource utilization, and the supported number of concurrent migrations. Consider the following when making your choice:

  1. Ultra Lite is least resource intensive than many of the options and supports cold migration using the PrimaryIO proprietary data transport.
  2. Lite mode requires the fewest number of resources among the options, but the number of parallel migrations and the number of concurrently migrated VMs will be limited.
  3. Standard mode is more scalable than Lite mode, yet less scalable than Performance mode. However, resource requirements will be proportionally higher or lower.
  4. Performance mode requires the highest number of resources among the options, but it also offers the greatest performance by delivering the highest number of parallel migrations and supporting the highest number of concurrently migrated VMs.

HDM Resource Requirements

The following table highlights the number of resources required for each HDM deployment type:

Note:

  1. Here the names ESXMgr, CloudCache, PremMgr and CloudMgr are abbreviated names of the VMs containing HDM components. These get created as part of the HDM deployment.
  2. The N in Cluster (N) refers to the number of nodes in the cloud cluster.
Standalone Cluster (N)
Ultra Lite
On-Premise 1 Appliance (4 vCPU, 8 GB RAM, 144 GB disk)

1 ESXMgr (8 vCPU, 8 GB RAM, 128 GB disk)

1 Appliance (4 vCPU, 8 GB RAM, 144 GB disk)

2 ESXMgr (8 vCPU, 8 GB RAM, 128 GB disk)

On-Cloud 1 CloudCache (6 vCPU, 12 GB RAM, 64 GB disk) 2 CloudCache (6 vCPU, 12 GB RAM, 64 GB disk)
Lite
On-Premise 1 Appliance (4 vCPU, 8 GB RAM, 144 GB disk), \ 1 ESXMgr (8 vCPU, 8 GB RAM, 96 GB disk) 1 Appliance \ 2 ESXMgr
On-Cloud 1 CloudCache (6 vCPU, 12 GB RAM, 32 GB disk, 512 GB cache) 2 CloudCache
Standard
On-Premise 1 Appliance (4 vCPU, 8 GB RAM, 144 GB disk), \ 1 PremMgr (3 vCPU, 4 GB RAM, 32 GB disk),

1 ESXMgr (4 vCPU, 4GB RAM, 160 GB disk)

1 Appliance

2 PremMgr,

N ESXMgr

On-Cloud 1 CloudMgr (4 vCPU, 6 GB RAM, 32 GB disk), \ 1 CloudCache (6 vCPU, 20 GB RAM, 32 GB disk, 512 GB cache) 2 CloudMgr, \ N CloudCache
Performance
On-Premise 1 Appliance (4 vCPU, 8 GB RAM, 144 GB disk) \ 1 PremMgr (3 vCPU, 4 GB RAM, 32 GB disk),

1 ESXMgr (5 vCPU, 9 GB RAM, 288 GB disk)

1 Appliance \ 2 PremMgr,

N ESXMgr

On-Cloud 1 CloudMgr (6 vCPU, 6 GB RAM, 32 GB disk), \ 1 CloudCache (8 vCPU, 40 GB RAM, 32 GB disk, 2 TB cache) 2 CloudMgr, \ N CloudCache

Recommendation: While any of the cluster modes can be used in the production environment, the Standard-Standalone mode is recommended for the test environment.