Dell NativeEdge: Guidelines for System Memory and CPU Reservations on a NativeEdge Endpoint

If Virtual Machine (VM) memory utilization or CPU utilization exceeds what is available to a VM, an Endpoint may become unresponsive and may reboot.

It is possible for a VM to utilize memory and CPU that is required for normal endpoint operations. If this occurs an out-of-memory scenario can occur, causing the Endpoint to become unresponsive and reboot. 

For NativeEdge 2.x, NativeEdge Engineering Guidance for Memory and CPU Reservation for Endpoints core services:

• RAM: Minimum 3 GB, recommended 4 GB

For NativeEdge 3.0, NativeEdge Engineering Guidance for Memory and CPU Reservation for Endpoints core services:

For nonclustered environment set aside:

• RAM: Minimum 3 GB, recommended 4 GB

For clustered environments set aside:

• CPU: For clustered environments, a general guideline is to allow for approximately 10% CPU overhead per node for system services. This value may scale with the number of datastores and IO workload intensity.
• RAM: Minimum 8 GB per node (including the 3 GB indicated for nonclustered environments) if a single local datastore is used for the shared storage datastore. For each additional local datastore that is added to the /Shared_Datastore, allocate an additional 4 GB per node.

For NativeEdge 3.1+, NativeEdge Engineering Guidance for Memory and CPU Reservation for Endpoints core services:

Additional checks have been implemented to prevent users from overcommitting endpoint memory.
(Note: This only applies to Virtual Machines (VMs) only, not container workloads)

For nonclustered environments set aside:

• RAM:  (Physical Memory in system) - 4 GB System overhead

For clustered environments set aside:

• CPU: For clustered environments, a general guideline is to allow for approximately 10% CPU overhead per node for system services. This value may scale with the number of datastores and IO workload intensity.
• RAM: See the table below for memory requirements:

Total memory assigned to all VMs on the endpoints depends on size:

A VM will not be assigned to an endpoint if the limit is exceeded.