Administrator Guide

Historical performance statistics for disks, pools, and tiers are displayed in graphs for ease of analysis. Historical statistics focus on disk
workload. You can view historical statistics to determine whether I/O is balanced across pools and to identify disks that are experiencing
errors or are performing poorly.
The system samples historical statistics for disks every quarter hour and retains these samples for 6 months. It samples statistics for pools
and tiers every 5 minutes and retains this data for one week but does not persist it across failover or power cycling. By default, the graphs
show the latest 100 data samples, but you can specify either a time range of samples to display or a count of samples to display. The
graphs can show a maximum of 100 samples.
If you specify a time range of samples to display, the system determines whether the number of samples in the time range exceeds the
number of samples that can be displayed (100), requiring aggregation. To determine this, the system divides the number of samples in the
specified time range by 100, giving a quotient and a remainder. If the quotient is 1, the 100 newest samples will be displayed. If the quotient
exceeds 1, each quotient number of newest samples will be aggregated into one sample for display. The remainder is the number of oldest
samples that will be excluded from display.
• Example 1: A 1-hour range includes 4 samples. 4 is less than 100 so all 4 samples are displayed.
• Example 2: A 30-hour range includes 120 samples. 120 divided by 100 gives a quotient of 1 and a remainder of 20. Therefore, the
newest 100 samples will be displayed and the oldest 20 samples will be excluded.
• Example 3: A 60-hour range includes 240 samples. 240 divided by 100 gives a quotient of 2 and a remainder of 40. Therefore, each
two newest samples will be aggregated into one sample for display and the oldest 40 samples will be excluded.
If aggregation is required, the system calculates values for the aggregated samples. For a count statistic (total data transferred, data read,
data written, total I/Os, number of reads, number of writes), the samples' values are added to produce the value of the aggregated
sample. For a rate statistic - total data throughput, read throughput, write throughput, total IOPS, read IOPS, write IOPS - the samples'
values are added and then are divided by their combined interval. The base unit for data throughput is bytes per second.
• Example 1: Two samples' number-of-reads values must be aggregated into one sample. If the value for sample 1 is 1060 and the value
for sample 2 is 2000 then the value of the aggregated sample is 3060.
• Example 2: Continuing from example 1, each sample's interval is 900 seconds so their combined interval is 1800 seconds. Their
aggregate read-IOPs value is their aggregate number of reads (3060) divided by their combined interval (1800 seconds), which is 1.7.
You can export historical performance statistics in CSV format to a file on the network for import into a spreadsheet or other application.
You can also reset current or historical statistics, which clears the retained data and continues to gather new samples.
For more information about performance statistics, see Viewing performance statistics, Updating historical statistics, Exporting historical
performance statistics, and Resetting performance statistics.
About firmware updates
Controller modules, expansion modules, and disk drives contain firmware that operate them. As newer firmware versions become available,
they may be installed at the factory or at a customer maintenance depot or they may be installed by storage-system administrators at
customer sites. For a dual-controller system, the following firmware-update scenarios are supported:
• The administrator installs a new firmware version in one controller and wants that version to be transferred to the partner controller.
• In a system that has been qualified with a specific firmware version, the administrator replaces one controller module and wants the
firmware version in the remaining controller to be transferred to the new controller (which might contain older or newer firmware).
When a controller module is installed into an enclosure at the factory, the enclosure midplane serial number and firmware-update
timestamp are recorded for each firmware component in controller flash memory, and will not be erased when the configuration is
changed or is reset to defaults. These two pieces of data are not present in controller modules that are not factory-installed and are used
as replacements.
Updating controller firmware with the Partner Firmware Update (PFU) option enabled will ensure that the same firmware version is
installed in both controller modules. PFU uses the following algorithm to determine which controller module will update its partner:
• If both controllers are running the same firmware version, no change is made.
• If the firmware in only one controller has the proper midplane serial number then the firmware, midplane serial number, and attributes
of that controller are transferred to the partner controller. Subsequently, the firmware update behavior for both controllers depends
on the system settings.
• If the firmware in both controllers has the proper midplane serial number then the firmware having the latest firmware-update
timestamp is transferred to the partner controller.
• If the firmware in neither controller has the proper midplane serial number, then the firmware version in controller A is transferred to
controller B.
NOTE:
Dell EMC recommends always updating controller firmware with the PFU option enabled unless otherwise
directed by Tech Support.
28 Getting started