Secondary Nmon Applications

Several applications exist for postprocesing nmon data. The first is called nmon_analyser [18]. The tool developed by IBM is an Excel spreadsheet that processes .nom files and outputs the results in tabs in a new spreadsheet. I won't cover all of the details in this article, but the tool is very easy to use. Just click on the Analyse nmon data button in the first spreadsheet tab, and it will prompt you for the input file.

The output spreadsheet contains plots, presented with the time series data, and metadata, presented as tabular data. I won't cover all of the tabs in the output spreadsheet because there can be many, but I will show two plots to give you an idea of what it produces.

The first chart in the spreadsheet is titled System Summary (Figure 10), and it presents two pieces of data. The first is the sum of the user and system CPU percentages, and the second is I/O transfers per second (roughly IOPS). Clicking through the spreadsheet tabs displays other data analyses, but the other plot I want to share shows percent CPU usage for user, system, and idle times (Figure  11).

Figure 10: Nmon analyser – CPU and I/O transfer plot for NPB FT example.

Figure 11: Nmon analyser – percent CPU usage for NPB FT example.

A Java-based tool named NMONVisualizer [19] can take nmon data from one or more CSV files and create plots. To use it, you just start the application and open the .nmon file; it then processes the data and creates a number of plots. The screen shot in Figure 12 shows a CPU chart of percent CPU usage by user, system, and wait time over the interval examined. The chart is only showing 80% CPU utilization because nmon continued to collect data after FT had finished running. All of the other tabs in Figure 12 contain other charts. For example, Figure 13 shows CPU usage for the system in the CPU over Time tab.

Figure 12: NMONVisualizer – percent CPU usage for NPB FT example over the entire interval.

Figure 13: Nmon analyser – percent CPU usage for NPB FT as a function of time.

Other secondary applications are available as well. If you use Python, you can analyze nmon output data via pyNmonAnalyzer [20]. The tool massages the nmon CSV data to make it more amenable to spreadsheets and creates two HTML-based reports.

One report creates static graphs, and the other has interactive graphs powered by dygraphs [21]. Unfortunately, I couldn't get pyNmonAnalyzer to work on my CentOS 6.6 box, even with the Software Collection installed; perhaps you will have better luck.

The nmon2rrd [22] tool converts the nmon CSV file to a website of graphs. I tried the tool, but it appears to work with the AIX version of nmon only. One last tool that might be useful is called nmon2web [23], which creates web pages from nmon data. I haven't tried this application, but it sounds pretty interesting.

Summary

I'm a big proponent of command-line tools to help debug misbehaving nodes, because many times I only have command-line access to the node. In the previous two articles, I covered two classes of tools I use: top-like tools and stat-like tools.

In my experience, the most common user issue is performance – Why isn't my application running? Why isn't my application running fast? Why isn't my application running as fast as it did yesterday? Therefore, I need tools to help me examine node performance, so I need a combination of top- and stat-like tools, which is why I like using nmon.

Nmon can be used as a top-like tool to watch what is happening on the system, but you can also use it to capture system information that can be used to understand what the system is doing. The many nmon options monitor what is happening on the system, and some even produce ASCII charts. It's a great way to get a quick visual impression of what is happening in the system.

Nmon can also capture a great deal of information from the system and produce CSV files for postprocessing. However, the results are typically not easy to postprocess, so you need to use one of the tools mentioned in this article to visualize the data.

Give nmon a chance. I think you will like what it can do for you.