Release Notes

18 Analyzing Dell PS Series Storage with SAN Headquarters | TR1050
10Gbps iSCSI networks are capable of accommodating typical bandwidth requirements. The PS Series architecture
effectively aggregates all the ports to address additional throughput needs.
2.4.3 Servers
The server should be eliminated as the source of a performance issue from a memory or processor perspective.
Although spinning disk access is much slower than memory access speeds, other factors that are non-disk-related still
need to be ruled out. For the highest availability and throughput, at least two NICs should be used. If Jumbo Frames are
enabled on the switch ports, the NICs should also have Jumbo Frames of the same size enabled. In addition, the
appropriate Host Integration Toolkits should be installed on the Servers attached to the PS Series SAN.
2.4.4 Tuning considerations
PS Series storage is a virtualized self-tuning system. The architecture is based on virtualizing all the disks and allocating
volumes over those disks from within a page structure. This allows for appropriate data movement to equalize capacity
and relieve performance constraints. To tune a PS Series system, a pool may need to be expanded by adding a new PS
Series array, or volumes may need to be moved to other pools within a group. This is a fundamental design concept to
the peer storage architecture where each member or peer will have its own disks, processor, memory, and iSCSI ports
to support the capacity and workload in that enclosure. When a new PS Series array is added to an existing pool, all of
the disks begin to have the data spread equally for all of the volumes in that pool. Those volumes then benefit from the
additional capacity and the IOPS capability brought to that pool.
2.5 SAN HQ GUI navigation technique
In SAN HQ, many views are available and may be overwhelming with the amount of functionality available. This section
describes a technique to help identify the problem quickly.
Below are some of the recommended views to help you determine the problem:
• Combined Graphs: Provides overall view of IOPS, MB/s, latency, and capacity trends. Can be used to observe
possible issues, such as high latency, high-bandwidth needs, capacity availability, and resource counts such as
iSCSI sessions.
• Hardware / Firmware: Provides overall hardware and software configuration. Is used to observe the number
of pools, members, and RAID preferences.
• Experimental Analysis: A tool which should be used as a guide to provide the theoretical maximum IOPS
based on the speed of drives, type of PS Series array, RAID preference, and read/write pattern. The
assumption for this calculation is 8KB I/O size. Always review latency and actual disk IOPS and queue depth
before making changes based on the Experimental Analysis tool.
• Disks: This shows IOPS at the individual disk level and typically indicates the source for high latency.