Black-Box + White-Box Analysis of RMF/SMF Performance Metrics
Project and Program: MVS
, MVS Performance
, SHARE Pittsburgh 2019
RMF (or CMF) and SMF contain by far the most robust set of infrastructure operations metrics from any computing environment in the data center. The metrics continue to grow with advances in the architecture for encryption, compression, etc.
The richer the metrics, the more value they contain not only for understanding performance problems in the infrastructure, but for quantifying the risk of performance problems in many different areas of the infrastructure. Most analysts recognize that this data source contains more value than they are able to effectively mine from it. This lack of easy visibility into the data leads to both performance problems and cost inefficiency.
For the deep experts with decades of experience, interpretation of the data is possible, but slower and less effective than it should be when they have to rely on static reports on the metrics. And as these experts retire, it is virtually impossible for new staff members to duplicate the level of expertise in a reasonable time frame which creates serious risk for the organization.
Fortunately, new algorithms to process the data are proving extremely powerful and practical to meet these needs for both the deepest experts and the new staff.
This session will examine the design philosophy of two approaches to solving this problem, and the pros and cons of each. Black-box analysis refers to the assessment of metrics using statistical approaches to recognize changes and anomalies in the data. White-box analysis refers to algorithms that assess and rate the metrics using information models built from digitized expert knowledge about the best practices and architecture specific. These modernized approaches to mining RMF/CMF and SMF for performance and efficiency will be discussed from a vendor-neutral perspective. Specific real-world user case study examples will be shown for both approaches.-Brent Phillips-Intellimagic
Back to Proceedings File Library