Outdated Version

You are viewing an older version of this section. View current production version.

SingleStore DB Studio Workload Monitoring

Info

SingleStore DB Studio is designed to work with MemSQL 6.5 or later and is only supported on Chrome and Firefox browsers at this time.

The SingleStore DB Studio Workload Monitoring page is a feature of SingleStore DB Studio that allows customers to record and visually analyze workload profiles at either the query or node level.

The Workload Monitoring page is built on top of the native Workload Profiling capabilities of MemSQL. As such, you should read the entire Workload Profiling topic before using Workload Monitoring in SingleStore DB Studio.

Creating a recording

To begin, you must create a recording of your query resource usage by clicking the Begin Profiling button in the UI. When you start recording a profile in SingleStore DB Studio, a snapshot of INFORMATION_SCHEMA.MV_ACTIVITIES_EXTENDED_CUMULATIVE is captured. When you stop recording, another snapshot of the same table is captured and SingleStore DB Studio calculates the difference between the two snapshots.

Analyzing a recording

SingleStore DB Studio allows you to analyze the resource usage of all the queries that were running in the cluster while the recording was taking place. For each query, you can inspect various metrics such as disk usage, network usage, and time spent on various things (waiting for disk I/O, waiting for CPU, etc.).

Moreover, SingleStore DB Studio also allows you to analyze the resource usage per MemSQL node (instead of per query). This is useful to see whether a specific MemSQL node is the bottleneck of a query’s performance.

Exporting and importing workload profiles

If you want to share a recording with other SingleStore DB Studio users, you can export a recording as a JSON file using the button in the upper-right corner of the page. Other users can then import these files through the Workload Monitoring page.

How does the Time Spent column work?

The Time Spent column on the Workload Monitoring page shows where a given query spent its time, as an average percentage of wall-clock execution time across all executions during the recording period. If advanced counters are not turned on, then this column will not yield the most precise results.