The Curious Case of Mainframe Re-hosting

Over the last few years, there have been several debates between the passionate mainframe experts community and CIO teams on the return on investment (ROI) brought on by re-hosting – whether it makes a sound business case to re-host mainframes to distributed workloads and whether distributed systems will be able to provide the scalability, availability and security that comes with EAL level 5 certified System z. In this article, we look at the nuances of migrating away from mainframe computing to distributed servers and the consequences it entails.

Enterprises need to look at various issues while making their decision, such as what runs best on which system and why, the key workloads that are best suited for a mainframe environment, etc. There is a need to look at the key workloads that exist within the enterprise and determine the value per transaction on a given platform – this is important, considering everyone expects the highest availability, highest security and seamlessly scalable systems.

Enterprises housing mainframe servers have been able to indulge in robust environments that come with a promise of 99.999% availability, virtualization and ease of provisioning capacity for additional workloads and, above all, an iron fisted security that is not so easy to hack.

With the rise of Linux and open source models, there have been various points of view on whether the spend on mainframe licensing is really worth it and if there are any alternatives - either to shift the work load to distributed systems or re-host the mainframe code on to a distributed environments in a bid to reduce the ongoing costs.

The debate around re-hosting has been primarily triggered by three factors, with the ostensible goal of reducing costs:

1. Ongoing licensing costs to be paid to IBM (News - Alert)
2. Availability of resources with mainframe skills
3. Adoption of newer workloads on the current mainframe platform

However, while these factors are relevant points for consideration, it is important to analyze whether the jump from mainframes to distributed servers will always lead to increased efficiency.  Over the last few years, mainframes have become more compatible with modern day platforms and still prove to be good bets in terms of transaction based computing.

So, the decision on re-hosting has to be evaluated in detail with emphasis on the following:

1. Types of workloads being run - The first step is to understand what the key workloads are -  whether I/O operations are high, whether the workload involves transaction processing or batch workload, etc. All these are very well run on a mainframe platform.  If the workload involves analytics and is CPU centric, then the organization needs to evaluate similar distributed platforms that can scale up and handle such workloads. The latest mainframes have specialty engines called System z Integrated Information Processor (News - Alert) or zIIP, System z Application Assist Processor or zAAP & IFL (Integrated Facility for Linux), that can help reduce cost by diverting either DB2 workload, Java application load or Linux workloads.

2. MIPS Optimization – Running a MIPS optimization program can yield potential savings, especially in the case of medium to large MIPS based environments. MIPS optimization can yield benefits in the form of reduced costs by reducing the consumption of computing resources, resulting in cost savings in the form of regular license/MSU charges being paid to IBM.

3. Assessment of Total Cost of Ownership– When evaluating the business case for re-hosting, a thorough analysis of complete cost of ownership of the environment is required. IBM does provide support in the form of IBM Eagle Study (ES) - this study gives you a head to head comparison of TCO of distributed servers with that of mainframe servers.

As Mainframes can be scaled up seamlessly, they require less power and space when the workload grows. In addition, IBM offers other optimization solutions of assessment such as IBM Scorpion from STG labs. These can also be investigated as part of TCO assessment.

4. Available Re-hosting Solutions & Risks of Migration – There are a number of re-hosting and modernization solutions available in the market, such as Microfocus or Dell (Clarity (News - Alert)). One has to carefully evaluate these products and determine if such solutions are compatible with current workloads. A thorough analysis of similar solutions, especially for DB2, IMS, CICS, can be initiated with a proof-of-concept to ensure these solutions work for your environment.

5. Benchmarking for Performance – Last but not the least, is the benchmarking on the performance and response times, since migrating workloads can result in sub optimal performances. A clear analysis and comparison of peak transaction workloads and response times for CICS or online transactions is necessary to determine the impact of migrating workloads . The data from this exercise can then be compared with the level of distributed resources required to meet similar performance requirements.

Re-hosting or re-platforming from a mainframe environment to a distributed environment is a big exercise. With the level of flexibility and upgrades available on mainframes, the decision to migrate to distributed environments may not always be the best idea for the business. Mainframes are one of the most secure and highly available environments, and so, a careful study considering business implications and risks needs to be considered.

About the author:

Girish Totloor is an Industry Principal with Infrastructure Management Services at Infosys (News - Alert). For more than 13 years he has been working on Mainframe Systems, Databases and applications performances for large enterprises. He has been advising clients across Retail, Logistics and Financial domain, focusing on deriving efficiency and optimization across workloads in the enterprise as a whole.