White Papers

8
Dell EMC Fault Resilient Memory
Introduction to Dell Fault Resilient Memory (FRM)
Figure 3 :Listing ESXi ReM kernel parameter
Commands such as esxtop and esxcli hardware memory get help the user to understand the memory
reserved for fault resiliency, when this option is enabled from sever BIOS.
The below screenshot is an output when you run the command Esxtop, and then press m. This output shows
the memory specific attributes. NUMA/MB field indicates the overall NUMA nodes created on the system
(when Node Interleaving is disabled in server BIOS). For example, the below screenshot is captured from a
Dell EMC PowerEdge R840 with FRM enabled and Node Interleaving disabled (NUMA Enabled).
Figure 4 :NUMA Representation from ESXi
In the above example, when Node Interleaving is disabled, there would be 4 NUMA nodes created in the
system each of 96 GB. With FRM enabled and Node Interleaving disabled together, first two NUMA Nodes
are reported with 48 GB of memory and last two socket NUMA nodes capacity is 96 GB each. This is
because 25 percentage of memory reservation to be allocated in the yx4x system and hence 96 GB (25
percentage of 384 GB) is reserved for memory fault resiliency. The 96 GB is further divided between first two
NUMA nodes equally to ensure that there is no NUMA node imbalance.
The command in the below screenshot depicts the reliable memory reported with in ESXi, as the same
example mentioned above. As noted above, system BIOS reserves approximately 96 GB as reliable memory
and hence 288 GB is reported as system memory. You may observe a slightly lower memory reported in
ESXi (287.2 GB) comparing the system memory reported in BIOS (288 GB). This is a known issue in ESXi
and is documented in VMware KB 2149889.
Figure 5 :Reliable Memory reported with in ESXi