Test lab compute and storage architecture
Test lab compute and storage architecture
-
This section provides information about the compute and storage configuration of the ax740sc101 cluster.
Compute configuration
The following table details the hardware inventory of each AX-740xd node. We configured each AX node with the same hardware in both sites as required. Only all-flash and all-NVMe storage configurations are currently validated and supported for stretch clustering on Dell Integrated System for Microsoft Azure Stack HCI.
Table 2. AX-740xd node configuration
Resources per AX-740xd
Description
CPU
2 x Intel Xeon Gold 6230R CPU @ 2.10 GHz, 26 cores
Memory
384 GB
Storage controller for operating system
BOSS-S1 adapter card
Physical drives for operating system
2 x Intel 240 GB M.2 SATA drives configured as RAID 1
Storage controller for Storage Spaces Direct (S2D)
HBA330 Mini
Physical drives for S2D
8 x 960 GB mixed-use Samsung SATA SSDs
Network adapters
- 1 x Intel X710 dual-port 10 GbE SFP+ and i350 dual-port 1 GbE, rNDC
- 2 x QLogic FastLinQ 41262 dual-port 10/25 GbE SFP28, PCIe full height
Operating system
Microsoft Azure Stack HCI, version 20H2
The following figure shows the nodes in the ax740sc101 cluster as shown in Windows Admin Center:
Figure 5. Servers view in Windows Admin Center
Storage configuration
We configured each site with a single storage pool. The following figures show how we created the volumes in Windows Admin Center:
Figure 6. Creating a volume in Windows Admin Center
Figure 7. Entering volume details in Windows Admin Center
For each data volume, we selected Replicate across two sites and set the replication mode to Synchronous. Because this cluster was set up as active/active, the replication direction could be either From Bangalore to Chennai or From Chennai to Bangalore, depending on which node the active VMs would be created. We created all volumes with two-way mirroring for the best mix of performance and resiliency.
Replication partnerships
The following tables summarize the replication partnerships that were relevant to the proof-of-concept test scenarios.
Table 3 shows the replication direction From Bangalore to Chennai. Table 4 shows the replication direction From Chennai to Bangalore.
Note: Some volumes, such as the Cluster Performance History volume and its related log files, are not listed in the tables.
Table 3. Replication direction: From Bangalore to Chennai
Bangalore (Site 1)
Chennai (Site 2)
VM disks
Volume name
Size
Volume name
Size
ax740xds1N1
800 GB
ax740xds1N1_Rep
—
10 VMs created using VMFleet
ax740xds1N1_Log
40 GB
ax740xds1N1_Rep_Log
40 GB
—
ax740xds1N2
800 GB
ax740xds1N2_Rep
—
10 VMs created using VMFleet
ax740xds1N2_Log
40 GB
ax740xds1N2_Rep_Log
40 GB
—
OM
100 GB
OM-Replica
—
Linux-based VM running Dell OpenManage Enterprise
OM-Log
40 GB
OM-Replica-Log
40 GB
—
Table 4. Replication direction: From Chennai to Bangalore
Chennai (Site 2)
Bangalore (Site 2)
VM disks
Volume name
Size
Volume name
Size
ax740xds2N1
800 GB
ax740xds2N1_Rep
—
10 VMs created using VMFleet
ax740xds2N1_Log
40 GB
ax740xds2N1_Rep_Log
40 GB
—
ax740xds2N2
800 GB
ax740xds2N2_Rep
—
10 VMs created using VMFleet
ax740xds2N2_Log
40 GB
ax740xds2N2_Rep_Log
40 GB
—
VMFleet virtual machines
We created 10 VMFleet VMs on each node for testing, using the following naming convention:
- vm-base-<node name>-<1-10>
The VMFleet VMs generated a moderate I/O load across the nodes in the cluster throughout the testing scenarios. In addition to the VMFleet VMs, we created a Linux-based virtual appliance named OME-1 on the cluster. OME-1 ran Dell OpenManage Enterprise software and was used to observe the behavior of a real-world application during the failure scenarios.