File System Throughput

Descriptions and sample graphs of File System Throughput view on Dashboard

The File System Throughput view aggregates live throughput performance metrics (reads and writes) from all active OSTs into a graphical format. In the graphs, file system throughput is tracked on the Y-axis with time (in hh:mm:ss format) shown on the x-axis. Read operations are shown in blue and write operations in green.

Show Averages Option

When selected, the Show Averages option displays average read and write data as curved lines on the file system throughput graph, enabling system administrators to quickly assess system performance trends. The following graph shows low performance.

To activate the average performance display, check the Show Averages check box in the legend at the bottom of the graph.
To deactivate the average performance display and return to the default detailed data display, clear the Show Averages check box.

Graph Zoom Options

The default file system throughput graph uses a one (1)-hour time scale on the X axis. The Zoom option allows system administrators to change the time scale to two (2) minutes, 10 minutes, and back to one (1) hour as needed when monitoring system performance.

To change the time scale, click the 2m, 10m, or 1h zoom options at the right side of the legend at the bottom of the graph.

Note: Read and write data is constantly refreshed at 5-second intervals, causing the graph to be updated with new data as it moves from right to left.

The following sample graphs show file system throughput data at different zoom levels.

Graphs With No Data or Interrupted Data

If read and write data is not available or if data gathering by the monitoring utility is interrupted, then the file system throughput graphs may look different than expected. For example, if Lustre is not operating for some reason, then no read/write operations will be taking place. In this situation, file system throughput data will not be available and an empty graph will be displayed.

Occasionally, system data gathering is interrupted, such as during an HA failover operation. As a result, no data can be graphed during this interval and a gap will appear in the graph (shown in the following graph).

Data gathering gaps also affect average performance graphs. If an insufficient amount of data is available to be reliably averaged, then the graph shows gaps for the period(s) of time when an accurate average could not be obtained. In these situations, the graph will look similar to the following:

Note: As soon as enough data is gathered continuously, the graphs will display normally.