In this digital transformation era of fast-paced and immensely scalable distributed applications, IBM Z Mainframes continue to be an essential player for major corporations. There is high demand for modern applications to access the data maintained in these operating centers. In their goal to build and deploy stable applications, DevOps teams require an understanding of how Mainframe services perform.
IBM Z Mainframes are complex and robust ecosystems that provide many data sources to analyze system health. System Management Facilities (SMF) is the prime and most reliable z/OS machine data producer for performance, workload activity, network, and security information. However, mining traditional SMF data for operational intelligence and data analytics is challenging when using the conventional method of recording and archiving these records, commonly referred to as the dataset mode approach or SYS1.MANx datasets. Some of the main problems of this method are:
- Managing and processing the enormous amount of SMF data records is an expensive and complex task
- DevOps organizations lack an easy way to visualize Mainframe information to satisfy their data management and real-time needs for just-in-time root cause analysis.
- Complex SMF recording and archival process.
- Data segregation, categorization, and scalability.
In z/OS 1.9, IBM launched the SMF capability to exploit System Logger services, offering many advantages compared to the traditional dataset mode, including:
- Performance: Data capturing is done faster, minimizing over buffer concerns when switching MANx datasets.
- Collection flexibility: Data can be grouped by merging multiple systems into a single log stream or by grouped and categorized based on the group’s needs.
- Backup: Logger manages a duplex copy of the data, which serves as a replica if the primary storage copy is lost.
- Data retention and dump. System Logger manages the data retention on an individual log stream. Thus, installations can determine how long the records in a particular log stream are kept, and when it is necessary, the utility IFASMFDL program can be used.
The SMF logstream mode has been available since 2007. Despite significant enhancements delivered since its initial release in z/OS 1.9, a recent survey that our organization conducted in some major enterprises in the United States, shows that a high percentage of companies have not made the jump to logstream mode, even though the initiative is in their project’s backlog. Contrary, we have found that many z/OS-based organizations In Europe have adopted this not-so-new approach.
The SMF log stream mode enables boundless possibilities to transform this data into valuable, actionable information based on metrics and traces for comprehensive real-time visibility and mainframe-inclusive observability.
Application Performance Monitoring (APM) and IBM Z Visibility
Many vendors offer capable full-stack APMs, but few offer mainframe support. And the few availble solutions come with a very high price tag. More than ever, DevOps team need to use these monitoring platforms throughout the development lifecycle for earlier debugging of problems and to gain confidence in achieving the expected service levels.
This lack of Mainframe observability impacts the DevOps team. When a span in a distributed application sits on the Mainframe side, a detailed representation of that operation and its attributes should disclose simple and concise data describing a specific problem, and where an increase in resource utilization can be associated with a decline in application stability.
Streaming near-real-time SMF data in logstreams, using a cost-effective middleware framework can facilitate the integration of IBM Z Mainframe Health data into any Application Performance Management Solution. Mainframe integration allows DevOps and other teams to ask the What, How and Why questions about mainframe application performance. In the following figure, we can see granular details of a DB2 deadlock in service span, by correlating SMF DB2 102 record.
Understanding mainframe-based applications is vital for the organization. Combining Mainframe calls together with other distributed services will help Mainframe-Backed and DevOps teams proactively identify and tackle a problem.
Exploiting the System Logger for the SMF data management and collection is an important step in implementing a modern streaming workflow to collect, transform, and integrate Mainframe data into your analytics toolchain.
Baran, O. & Sica, A. (2008). Migrating SMF from Data Set Recording to Log Stream Logging. [White paper]. IBM.
Kyne, F., Dagmar, F., Girona, J., Klaey, W., Lundgren, L.. (2011). SMF Logstream Mode Optimizing the New Paradigm. IBM Redbook.
Heinrich, A. (2021). DevOps: A Modern Industry Revolution. Enterprise Tech Journal.
Sabharwal, N & Shukla, R (2022). Application Observability with Elastic. BPB Publications.