• Monitor system performance using Unisphere

    PDF

    Monitor system performance using Unisphere

    About performance metrics

    You can analyze system performance by viewing and interacting with charts that display historical performance data. These charts enable you to:

    • Compare changes in performance across multiple metrics, such as network traffic, bandwidth, and throughput.
    • Analyze data at both the aggregate and detail levels:
      • Line charts initially display aggregated performance metrics information, enabling you to quickly determine whether there are any high-level performance issues.
      • Once you open a line chart, you can drill down to the detailed data by applying breakdown categories to the displayed data.
    • Use preset or customized time ranges to view data collected within a specified time period.
    • Export the data of selected metrics to a .csv file. The exported values reflect the metric's filtered state.

    Metrics collection is enabled by default. The system automatically allocates space for storing collected metrics, whether or not metrics collection is enabled.

    About line charts

    A line chart initially displays aggregated information for one performance metric, such as LUN Bandwidth. The line chart shows the metric's value along the y-axis against the user-specified time range on the x-axis. When you hover over a data point on a line chart, a tool tip displays the time, date, and measurement associated with that data point.

    Line charts provide metrics for the following resources:

    • System cache
    • System I/O, including SMB (CIFS) and NFS file systems
    • System resources, including CPU utilization
    • LUNs
    • File systems
    • Fibre channel ports (physical deployments only)
    • iSCSI interfaces
    • Ethernet ports
    • Disks

    System line charts display all information across all like resources in the storage system. For example, the System-IO File System Bandwidth metric displays the total amount of file system I/O requests, in KB/s, across all file systems in the storage system. Resource line charts display all information for the selected resource objects. For example, the File System Bandwidth metric displays the total amount of file system I/O requests, in KB/s, for the selected file systems.

    Depending on the metric, each line chart aggregates data using one of the following calculations:

    Table 1. Calculations used for aggregating data
    Calculation
    Description
    Sum
    Total of the measurements for each target object instance. For example, the LUN Bandwidth metric adds together the amount of I/O traffic handled by each LUN.
    Average
    Sum of a set of numerical measurements divided by the number of measurements in the set. For example, the system uses the following formula to calculate the CPU Utilization metric, which is calculated as a percentage:
                                  [(Percentage of processing cycles used by SP1) + (Percentage of processing cycles used by SP2) / 2]
                                
    Weighted average
    Average that gives more weight to the objects with the highest contribution to the metric. The storage system uses the following formula to calculate a weighted average:
                                  [(m1 * w1) + (m2 * w2) + (m3 * w3) / (w1 + w2 + w3)] 
                                
    where:
    • m1 is the metric for object 1, m2 is the metric for object 2, and so forth.
    • w1 is the weight for object 1, w2 is the metric for object 2, and so forth.
    For example, the storage system uses the following formula to calculate the LUN IO Size metric:
                                  [((LUN1 I/O size) * (Number of LUN1 I/Os)) + ((LUN2 I/O size * (Number of LUN2 I/Os)) + ((LUN3 I/O size) * (Number of LUN3 I/Os))/ (Number of LUN1 I/Os) + (Number of LUN2 I/Os) + (Number of LUN3 I/Os)]
                                
    In this example, the LUNs with the highest number of I/Os have the highest weighting.
    Ratio
    Relative size of two values. For example, the FAST Cache Dirty Ratio metric measures the ratio of dirty pages to total pages in the FAST Cache.

    For information about the calculation used by a particular metric, see the help topic for that metric.

    Line charts typically display 150 - 300 samples. Within a line chart, different time ranges correlate to specific sampling intervals, as shown in the following table:

    Table 2. Time ranges and associated sampling intervals
    Time range setting
    Sampling interval
    3 Hours
    1 minute
    1 Day
    5 minutes
    1 Week
    1 hour
    Custom
    Varies based on the length of the range. If the range is more than one week, the sampling interval is 4 hours.

    If you select a preset time range for a metric (such as 1 Week), but the amount of available data is less than the associated display interval for that preset, the system tries to display the data using the next smallest available interval. This behavior can also occur when you page back and forth through time.

    For example, suppose you select the Cache Read Hit IOPS metric and the 1 Week preset. The system tries to display the metric data in one hour intervals. If there is less than one hour's worth of metric data, but more than five minutes' worth of metric data, the system displays the data using five minute intervals. If there is less than five minutes' worth of metric data, but more than one minute's worth of metric data, the system displays the data using one minute intervals. If there is less than one minute's worth of metric data, the system defaults to the most granular sample size possible, while keeping the legibility of the graph intact.

    Performance metrics aging policy

    The system governs the aging of collected metrics data according to the following goals:

    • The total consumed space for storing metrics should not exceed 70% of the available metrics space (which is 16 G).
    • Metrics data retention is based on the sampling interval:
      Table 3. Sampling intervals
      Sampling interval
      Retention period
      1 minute
      3 days
      5 minutes
      14 days
      1 hour
      28 days
      4 hours
      90 days

    If the metrics retention policy is violated due to lack of space, the system decreases the retention period for the oldest metrics, while attempting to retain at least 24 hours worth of data for all collected metrics. The system chooses the smallest sampling interval that can satisfy this goal without violating the first goal.

    View aggregated metrics data

    To view and analyze aggregated performance metrics, you use a line chart. A line chart displays aggregated information for one performance metric, such as LUN Bandwidth. It can help you identify trends in the performance metrics data.
    Procedure
    1. Under System, select Performance.
    2. Select the dashboard for the system for which you created a performance metrics display.
    3. A performance metrics line chart displays a summary or system level view of all the performance metrics from the system objects being monitored such as disks and LUNs.
    4. For each system dashboard, you can define the time range of the values displayed for all the metric line charts on that dashboard. The default time range is All. Alternatively, select one of the other default time range values (15 minutes, 3 hours, 1 day, 1 week).
      The default time range selections are enabled only if Unisphere has data spanning that time range.
    5. Select Custom or use the navigator bar to select the start and end dates and times of the values displayed. Select Update to get the latest data.
    6. To drill down into the data displayed in the line chart, you can breakdown the data displayed into individual lines that show the categories and contributors that provide data to the performance metric. Choose among the breakdown categories available for a particular metric.
      Each contributor displays as a different color line in the chart and is identified in the legend. You can quickly remove and add each contributor by clicking on its name in the legend. Use the breakdown display to determine if one contributor is adding to the aggregated total more than another contributor as well as analyze how a contributor's activity increases or decreases at a particular time.
    7. Hover over a data point in the chart to display the date, time, and measurement associated with that data point. Gaps in metric data collection are displayed as gaps in the line chart.

    Add charts

    Use Add Charts to select the performance metric charts to be displayed on each dashboard tab.
    Procedure
    1. Under System, select Performance and then select Add Charts.
    2. Select the resource type for which you want to display metrics.
    3. From the list of available metrics for that resource type, select the metrics you want to display. If applicable, select the objects for which you want to display metrics.
    4. Select Generate Charts to save your selections. You must select Generate Charts before you can select metrics from a different resource type.
    5. Select another resource type and metrics.
    6. When you have completed your selections, select Close.
    Results
    There is no limit to the number of metric charts you can display on each dashboard tab. Newly added charts display at the top of the display; you can drag and drop charts to different positions.