With the --docker-metrics option, NoSQLBench downloads and launches containers to support Grafana. You will know when Grafana is running by watching the log message on the terminal.

Wait for the log message telling you to go to port 3000.

Then, click here to change to the Grafana tab and log in. The credentials are:

• username: admin
• password: admin

Normally, you should change your password when you login the first time, but for this training scenario, just skip it.

Next, click on the Home icon.

Then, click on NoSQLBench.

The benchmark will need to run for a minute or so to allow statistics to accumulate. Then you will start to see results in the various metrics' graphs.

This is an easy way to grasp the performance of your benchmark!

Let's look at what the various metrics represent.

The top of Grafana's display is an explanation of the various metrics.

Notice the Ops and Successful Ops graph on the left of Grafana's display. This graph indicates the operation rate (reads/writes per second) of the benchmark. Use this metric to see the throughput achieved by the benchmark in this configuration.

The next graph shows how many database errors have occurred. You probably won't see any errors, but if you do see errors, they may be an indication of database stress.

The third graph indicates the mean service times for operations based on percentiles. Use this metric to understand how quickly the database responds to requests.

The fourth graph on the top row shows how many attempts (tries and retries) the benchmark performs per operation. Normally, each operation will only use one try to complete the operation. However, as databases become swamped, timeouts will cause the try count to increase.

There are additional metrics which you can read about, but initially, the first four are the most useful metrics.

Wow! Grafana gives you a great view of effects of the benchmark!

Grafana actually pulls its data from Prometheus, which gets the data indirectly from NoSQLBench. You can get to the Prometheus tab by clicking here .

Prometheus is useful if you are a die-hard metrics junkie who wants the raw footage. For example, if we want to investigate the raw one-minute mean service time data, we would follow these steps.

First, Select insert metric at cursor and select the metric.

Click the Execute button.

Select the Graph tab.

You should see something like this.

Now, you can go spelunking through as much raw data as you want!