Mainframe Monitoring: Beyond the Black Box
This post is also available in: Spanish
Mainframe Monitoring: Introduction, Challenges and Tools
Either because we are implementing an integral monitoring platform, because we want to monitor a data centre or because we need to monitor an application, sooner or later we will have to face the issue of Mainframe Monitoring.
There is a part of the IT world that thinks that factors such as the costs associated with mainframes, the development of servers of other architectures, virtualization and the consolidation of technologies associated with the cloud have sentenced the mainframes to disappear.
However, the Mainframe world has statistics that support the current relevance of this type of equipment. CA Technologies has been handling the following data since 2016:
- 55% of business applications need a Mainframe.
- A Mainframe accounts for 70% of business transactions.
- Between 70% and 80% of corporate data is on a Mainframe.
On the other hand, Compuware published in 2015 its document Mainframe Excellence in which it states that:
- By the year of publication, more than 220 billion lines of code were used in systems associated with Mainframes, and they estimated an increase of five billion each year thereafter.
- 87% of companies that used Mainframe by 2015 expected to continue to do so in the next five years.
There is another belief contrary to Mainframes; we tend to relate the presence of Mainframes with large companies such as banks, insurance companies, government organizations, etc..
However, in 2017 IBM launched two new models of so-called “thin” servers.
The idea is to offer SMEs servers that start from the characteristics and advantages of a Mainframe z/OS, but in a smaller architecture that is better related to technological trends (use of the LinuxOne operating system, equipment ready for the cloud and suitable for machine learning, etc.).
In any case, whatever the future of Mainframes, if we already have one of these servers within the structure we want to monitor, we must consider Mainframe Monitoring.
Mainframe Monitoring Challenges
Monitoring a Mainframe brings a set of challenges among which we can highlight:
- The traceability of an application: The idea of monitoring the entire user experience and detecting performance bottlenecks can be difficult to implement if our application has a service provided by a Mainframe, for example if it is a WEB application that uses Mainframe services as a back-end server.
This difficulty has led to the quickest solution, to consider this service as a black box, so we can have information on the total time it takes to resolve a transaction in the mainframe, but it does not say anything about how the transaction moves within this type of servers, thus losing analysis and optimization capacity.
- Specialized technical personnel: Generally, personnel with technical expertise in Mainframes are not the same as personnel with expertise in servers of other architectures, which often complicates the monitoring and optimization activities of the platform.
In addition, in recent years companies such as IBM, its customers and associates report problems to achieve the renewal of technical staff specialized in mainframe, so there is a possible structural shortage of qualified personnel.
- Lots of tools: The mainframe monitoring market presents many product options, with different capabilities, implementation methods and purchase and payment schemes (many products are purchased for a license fee but involve payments per processing cycle, for example).
Choosing the right tool and achieving the knowledge needed to extract performance from it can be a complex job, particularly as we will depend on the knowledge and skills of the mainframe specialist.
On the other hand, an integral console that allows the monitoring of all the resources of the platform, including the present mainframe(s), becomes a challenge associated with the presence of so many tools.
How do we do Mainframe Monitoring?
Bearing in mind the challenges it is time to face the big question: how do we monitor Mainframe z/OS servers?
A first approach may be to review the tools specially designed for this type of servers.
Evaluating mainframe environment tools
Specialist Robert Crawford distinguishes three types of monitoring tools for Z/OS mainframes according to the time window in which the tool is effective:
- Real time monitors: They allow us to visualize in real time variables such as amount of memory or transactions on the input/output subsystem.
The monitors in real time are usually programmed in function of thresholds, so that at the moment of surpassing this threshold the capture of data begins.
They have two main disadvantages: firstly, the overload that they can suppose for the Mainframe and secondly, that given their immediacy they do not allow to infer the behaviour in the long term.
- Recent time monitors: We can understand these types of tools as the mid-point between real-time visibility and historical data analysis.
So a recent time tool will accumulate data on the subsystem we want to monitor in fixed time periods, which are separated by fixed non capture intervals.
For example, IBM’s Resource Measurement Facitility Monitor displays data for periods of no less than 60 seconds with no pick-up intervals of 60 seconds or more.
Then, the tool will allow us to visualize and work all the data already collected, so the overload generated by this type of tool is much less than real time tools.
The main disadvantage is that by defining noncatch intervals they can overlook events that could affect performance.
- Post-processing monitors: These tools that accumulate a large amount of data to make it visible, usually the next day, are usually more related to the analysis needed to make a capacity plan than to a monitoring scheme.
Among its disadvantages, besides the cost in resources and licensing, it is the fact that reviewing and analysing the data coming from this type of tools can be a very hard task and requires a lot of knowledge of the details about the mainframe subsystems.
Of course there are tools that perform both real-time monitoring and post-processing; in this category the Omegamon product suite deserves special reference.
Omegamon products began to be developed as third party tools and after IBM acquired the company, they are marketed as part of the IBM Tivoli family.
Omegamon is interesting because it offers multiple options that fit the characteristics of the mainframe and because it manages to consolidate in a single console all the monitoring data of the server allowing to isolate problems and to take action.
Evaluating integration with general purpose monitoring tools
The next step to achieve Mainframe Monitoring can be to evaluate the use of general purpose monitoring tools (networks, applications, servers, etc.) in order to integrate mainframe monitoring data into this comprehensive visibility.
This integration would have the additional advantage that it would transfer the mainframe information to a group of operators who might not be mainframe specialists but with the right information and the right procedures could perform level 1 troubleshooting and support activities, and then bridge the gap with z/OS environment specialists if necessary.
There are two options that we think are worth highlighting. One of them is the IBM z/OS environment tools called ASG-TMON, which present a range similar to Omegamon but integrated with TCP/IP platform monitoring tools and applications.
The other solution is the one achieved by Pandora FMS with its product Pandora FMS for z/OS, which allows monitoring in real time of the mainframe resources without the need to install any additional tool in the server.
Pandora FMS solution for z/OS is, without a doubt, a good example of integration of the mainframe monitoring data generated by a standard component such as the MDF Dataportal Monitor to a general purpose monitoring tool such as Pandora FMS.
Our readers can check a practical application of the Pandora FMS solution for z/OS to the area of cost savings associated with the Mainframe in this publication of this blog.
Without a doubt, the topic of Mainframe Monitoring is interesting and complex because it depends on the tools you have on your server, the monitoring tools you use for the rest of your platform, the knowledge about mainframes you have, your expectations and of course your budget.
We hope this article will serve as an introduction to the subject and we invite you to tell us your experiences and to request more information about Pandora FMS for Mainframe Monitoring in this link.