Starting with release 4.2.1 IBM Spectrum Scale users have the option to allow 3rd Party performance visualization tools to access the internal Spectrum Scale performance data. This functionality is provided by the IBM Spectrum Scale Monitoring Bridge. The user can create their own performance graphs using Grafana dashboard with the IBM Spectrum Scale Monitoring Bridge.

Introducing IBM Spectrum Scale Monitoring Bridge

The IBM Spectrum Scale Performance Monitoring Bridge is a standalone Python application developed by the IBM Research - Zurich team. It translates the IBM Spectrum Scale metadata and performance data collected by the IBM Spectrum Scale Performance Monitoring tool to the data organizations and query requests used by openTSDB. The data exchange mechanism, via TSDB approach, uses the key-value pairs schema instead of ‘classic’ metric names in the queries. This has several advantages e.g. it makes the metrics more self-descriptive and allows creating data dimensions (tags) without changing the metric identity.  In the classic approaches like DB2 such changes would break existing data displays.

Figure 1: IBM Spectrum Scale integration framework for Grafana

The BM Spectrum Scale Performance Monitoring Bridge is easy to set-up and requires small resources overhead. It acts as the datasource for Grafana dashboards and provides the full set of IBM Spectrum Scale performance metrics.

Why the IBM Spectrum Scale Performance Monitoring Tool was developed for using with Grafana?

Grafana is an open source Graphite dashboard editor for large-scale data displays on demand. It is excellent at creating real-time graphs of numeric and time-differentiated data. It contains several more options for flexible dashboard design and navigation. It enables easy metric and function editing. As result, the user can create a scalable monitoring system for visualization performance data from more than one data source over long time ranges. Looking at Graphite’s graphs he can diagnose certain issues without diving into logs. However, Grafana is a tool for data presentation and provides limited functions for business analytics or system management functions such as historical data analysis, system configuration or status view. In comparison to another open source visualization tools, like Kibana, which works on top of the Elasticsearch, Grafana is not limited to any specific data source. Grafana doesn't have any local data store or log shipping.

Grafana allows the user to combine data from multiple data sources into a single Dashboard. At the same time the data from a specific Data Source, which belongs to a particular Organization, could be grouped  in to Panels.

Figure 2: Dashboard data organization example

The Tags filter provides additional option for data search or internal structuring.

Figure 3: Example of a dashboards grouping by using “tags”

One of the most powerful and most useful features of Grafana is the Dashboard Templating. By using template variables instead of hard-coded parameters in the data queries, dashboards can be made generic.

Figure 4: Example of a variable, making the node name selection flexible ($nodeName)

Figure 5: Example showing the dashboard after replacing the node name with the variable $nodeName in the query filter

Through the Export/Import function any dashboard can be distributed and used in different environments.

Figure 6: Importing predefined dashboards into Grafana

Grafana provides a number of ways to share the data with other users. Either a whole dashboard including current selected time range and template variables, or just a snapshot of the specific graph without sensitive data, like query or template variable, can be made accessible for others.

Figure 7: Data sharing options

For more information about graphite powerful features please check the Grafana online documentation.

Which software can be used for IBM Spectrum Scale Performance Monitoring alternatively

IBM Spectrum Scale Performance Monitoring GUI

A good start to IBM Spectrum Scale performance monitoring is to just use the IBM Spectrum Scale GUI dashboard. The GUI provides an easy to read and real-time user interface that displays a graphical representation of the status and historical trends of key performance indicators. This helps the users to make decisions easily without wasting time evaluating CLI data queries output or installing and configuring additional software tools. However, the user of the IBM Spectrum Scale Performance Monitoring GUI has limited flexibility in a dashboard customizing and adaptation of the individual operational needs.

Performance monitoring with IBM Spectrum Control

A powerful framework for performance data evaluation, threshold violation and reporting can be found in IBM Spectrum Control, an enterprise solution for SAN wide environment management. It collects data from all devices found in the local storage area using the devices native interfaces. The usage of the IBM Spectrum Control is good approach for monitoring a complicated storage network environment, predict warning signs of system failure or doing capacity planning as overall workload grows. The IBM Spectrum Control performance monitoring is focused on I/O traffic representation splitted in front-end (traffic between servers and the storage subsystems) and back-end (traffic between the subsystem cache and the disks in the RAID arrays). For the IBM Spectrum Scale environment IBM Spectrum Control monitors only CPU, Memory, GPFSNode and GPFSFileSystem metrics.

I don't think there is an answer for what's a better tool. It depends on the customer environment. The first step to choose which of these tools to use is the conclusion how big is the environment to be monitored, just a single IBM Spectrum Scale cluster, more than one cluster or a SAN wide solution is preferred. The second step is to determine the importance of analytics to the user, vs more visualization-only tools. If analytics is a big driver, probably the capacity estimation for historical data allocation and archiving need to be performed. Finally, the decision would probably depend on the available budget. The downside to open source tools is they require the user to install and set them up by himself. For log management tools keep in mind the implicit cost of the used bandwidth. In most cases log analyzers tend to use much more bandwidth than metrics tools. Conclusion: IBM Spectrum Scale monitoring bridge for Grafana is an easy way to visualize and share performance data of the IBM Spectrum Scale environment, without diving into more complicated analytics or data collection.

The installation guide and the download package of the IBM Spectrum Scale Performance Monitoring Bridge can be found at the IBM Spectrum Scale wiki page: IBM Spectrum Scale Performance Monitoring Bridge.