For the Ready Stack SAN design, consider the following recommendations:
- For SAN arrays, always design for two FC fabrics to ensure high availability connectivity for the storage array.
- Base the number of port connections, whether SAN or IP, on throughput requirements and tolerance for degraded mode operation during a path failure.
- When using vSphere:
- Multiple datastores within the vSphere cluster enable the use of vSphere HA Datastore Heartbeat. Enabling datastore heartbeating ensures that partitions or isolated host networks do not trigger VM movement within the cluster.
- VMware currently supports a maximum datastore size of 64 TB and 2,048 powered-on VMs per VMFS datastore. However, in most circumstances and environments, a target of 15 to 25 VMs for each 500 GB to 750 GB datastore is the conservative recommendation. LUNs and VMware vSphere datastores can be easily expanded to address future growth. Maintaining a smaller number of VMs per datastore greatly reduces the potential for I/O contention, which results in more consistent performance across the Ready Stack environment.
- Using thin provisioning within VMware on virtual disks does not initially result in additional space efficiency when thin provisioning is enabled on the array. However, the ability to reclaim space from within a compatible guest operating system requires that thin provisioning be used on both the storage and the virtual disks.
- Determine the appropriate disk size for your workload. The dynamically expanding disk type works well for most workloads on Unity All-Flash. Because Unity arrays use thin provisioning, only data that is written to a virtual hard disk consumes space on the array, regardless of the disk type—fixed, dynamic, or differencing. Therefore, the best disk type to choose more often depends on the workload rather than how it will impact storage utilization. For workloads generating I/O, such as Microsoft SQL Server databases, Microsoft recommends using the fixed-size virtual hard disk type for optimal performance.
- Use separate storage LUNs for data and log files. To rule out disk contention, Dell Technologies recommends using separate LUNs. For best performance, create distinct LUNs for a server’s datafiles and log files.