Storage Performance Reports > Performance Profiles Overview
Version 9.2.01
Performance Profiles Overview
Data Collection of array performance data is supported for certain vendor storage arrays. See the APTARE StorageConsole Certified Configurations Guide for a complete list of supported arrays.
Performance profiles reflect key performance metrics for groups of LUNs. These profiles are used to assign performance states to LUNs and hosts, as shown in several APTARE StorageConsole array performance reports. APTARE StorageConsole maintains aggregated performance profiles for each collection of RAID groups and thin pools that share common physical characteristics. These performance profiles categorize a host’s performance state, based on LUN performance: Slow, Normal, or Fast. All LUNs are mapped to either a RAID group or thin pool thereby enabling array performance reports to list the associated drives, drive types, and drive speed. See the following reports: Array Performance by RAID Group and Array Performance by LUN.
Performance profiles are derived from previously captured array raw performance statistics. APTARE StorageConsole’s internal procedures examine captured performance metrics for the specified interval and populate profiles with actual ranges of performance and throughput metrics for slow, normal, and fast performance states.
Performance profiles should be based on a typical peak array performance interval. By default, the updating of performance profiles occurs daily and uses an interval of 10:00 a.m. to 11:00 a.m. You should change the default interval to correspond to the peak array performance interval for your environment. When the Data Collector policy is configured, a recurring frequency is selected to set the length of the profile interval. Baseline profiling determines the range of metrics for I/Os, response time, and cache hits. See Populating Performance States.
An on-going scheduled daily task profiles collected data. Since certain times of day are more prone to performance issues, the performance profile schedule can be modified for your environment. For Performance Profile Schedule Customization, see Customize the Array Performance Profile Schedule in the System Administrator’s Guide.
Both internal and community profiling can be enabled.
Internal profiling compares periodic collection cycles within just your environment. Essentially, a library of metrics is maintained for comparisons.
Community profiling lets contributors securely import/export performance profiles over https, enabling comparisons with a broader community of similar arrays and environments. See Community Performance Profiling and Aggregated Profiles.
Determining Performance States
After performance profiles have been created, future reports will use the performance profiles for determining the performance state of LUNs, hosts, and other objects (in future releases).
Populating Performance States
Performance profiling establishes criteria for deriving a LUN’s performance state, as part of future APTARE StorageConsole reporting. The criteria is established based on actual performance metrics captured from each LUN. Reads and writes are treated independently.
Updating Performance Profiles
Performance profiles are used to characterize response times, I/Os per second, MBs per second, and cache hits per second as Poor, Normal, and Fast. Creation of and updating Performance Profiles establishes the criteria used based on performance and throughput metrics captured within a profile interval, using the following logic:
1. For all LUNs with performance data captured during the specified performance interval, group the data based on common physical attributes: array vendor, array family, array model, RAID type, number of drives in RAID Group, drive type, drive form, and drive speed. Each unique grouping of LUNs based on this criteria represents an individual APTARE StorageConsole Performance reporting profile entry for this Portal.
2. Calculate the average and standard deviation for all captured metrics based on all LUNs with performance data in the specified interval and within a common performance profile. When there is more than one poll to the array during the profile updating interval, averages and standard deviation will be based on only the portion of the interval covered for the polls that return data, ignoring polls that did not capture I/Os or where the response times were zero.
3. To determine boundaries within the range, subtract one standard deviation from the average. If the standard deviation is larger than the average, divide the average by two. This establishes the low end of the Normal boundary and the high end of the Fast boundary.
4. Add one standard deviation to the average. If the standard deviation is larger than the average, add half of the average to the average. This establishes the high end of the Normal boundary and the low end of the Poor boundary.