ISG Provider Lens™ Mainframe Services & Solutions - U.S. 2021 - Mainframe Operations

This is the first time ISG Provider Lens™ studies have included mainframe modernization. It is a response to buyers' requests for advice on what to do with their mainframes. This research qualified 47 vendors and service providers in five quadrants. According to these companies, the mainframe modernization market has been accelerating in the last two years, driven by the need to increase business agility.

In the preparation phase for this study, we found a lot of articles speculating about what a CIO should do with the mainframes. Common questions include: How to include the mainframe in the cloud strategy? As COBOL programmers are getting close to retirement, what are the risks of facing a skill shortage?

In this report, clients will find five alternatives represented in each quadrant. Mainframe modernization considers the choice of introducing agility into legacy mainframe applications. Mainframe transformation deals with options to move all applications off the mainframe. Mainframe-as-a-Service (MFaaS) supplements the modernization path, providing a pay-as-you-go (PAYG) business model. Those that do not plan to modernize consider outsourcing mainframe operations. These four quadrants help clients find the right service provider that can deliver to their needs. The fifth quadrant provides clients with modernization tool options for those that prefer to run modernization projects themselves.

A CIO should reflect on the real issue before figuring out if there is a need to modernize or change the mainframe. Setting a short timeframe (12-18 months) is imperative to guide reasoning in this case. More than 24 months would lead to a bias in favor of long application modernizations, which are not recommended. In this report, decision-makers will find more details about the key topics that follow.

Mainframe legacy applications are dead: IBM Z platform has a future running z/OS, Linux, and other operating systems. However, for enterprise clients, the future of the hardware is irrelevant. Business applications are what is important. COBOL, Assembler, PL1, Natural, and other legacy programming languages are procedural and outdated. Modern applications are built on object-oriented programming languages such as Java, .NET, and C#. Investing in modernizing the toolset for agility will bring more significant benefits in the long run.

The top 100 have mainframes: According to IBM, “92 of the world's top 100 banks, 23 of the 25 top U.S. retailers, and nine out of 10 of the world's largest insurance companies run System z mainframe. Nine out of the top 10 global life and health insurance providers use a System z mainframe. And 71 percent of global Fortune 500 companies are System z clients.” These facts do not suggest that enterprises should be writing new COBOL applications.

Optimizing the mainframe: Clients have several tools to optimize their mainframes. Providers of MFaaS and Mainframe Modernization can bring 10 to 25 percent cost savings while transitioning services. They check configurations, software licenses, and dead code that may waste mainframe resources. Also, competitive bidding processes have historically helped reduce cost.

Modernizing the mainframe: The most frequently used method is encapsulating Batch and business functions into microservices that can run directly in the cloud. Application programming interfaces (APIs) that expose business function or mainframe data to other applications are always cited as part of the modernization. When the intention is to decommission the mainframe, modernization results in moving all applications to the cloud (i.e., replatforming). Two methods that are prevalent include emulators, which enable COBOL to run in the cloud, and COBOL code compiled to Java or .NET to run in the cloud.

Re-engineering applications off the mainframe: A method that is gaining momentum is automated application re-engineering. Tools are fast, reliable, and produce quality code. Recent advances in methods and technology include artificial intelligence (AI), programming frameworks, code quality inspection, and automated testing. In the past, these tools required expensive on-premises processing power; however, at present, these tools run in the cloud with an increased processing capacity at much lower costs as well as much lower risks. Re-engineering of applications is viable and cost-effective.

Re-engineering: Most case studies are under five million lines of code converted to Java. Other languages include .NET and C#. Re-engineering is completed in a few months. The largest case study was 20 million lines of code converted in 20 months. Automated reengineering can convert two million lines of code in one  hour. Most of the project duration is spent on testing and quality assurance.

Converting COBOL to Java: Direct conversion does not include re-engineering. Data and logic stay the same, and the new code behaves the same as the old code. These conversion tools handle COBOL and many legacy languages and write modern code where Java is the most popular language. Converting code is much faster than re-engineering but also involves many testing cycles. These automation tools can convert 28 million lines of code in one hour.

Emulators: Replatforming and moving applications to x86 servers from mainframes have long been a possibility. The recent development observed is cloud virtual machines have increased the capacity of each x86 server, and the virtual servers can scale horizontally (many server images install within minutes). Cloud capacity and improved emulation technologies enable workloads of more than 100,000 MIPS to run in the cloud.

The database and storage myth: Mainframes hold vast amounts of data, which suggests that mainframe databases cannot go into the cloud. However, none of the participants in this survey mention issues associated with database size or storage complexity. Cloud data lakes are popular alternatives for storage, flat files, and virtual tape backup. Service providers are unanimous in converting legacy databases to any relational database with automated tools. The most popular choice to replace IBM DB2 is the open-source PostgreSQL.

The performance myth: Mainframes scale vertically, by adding more disk, CPU, and memory. The cloud scales horizontally, adding more servers of the same capacity. Any of the methods for replatforming mainframes to the cloud offer the same performance, or
better, because of horizontal scaling.

The skill shortage myth: Service providers have demonstrated they can attract and train young talents to work on mainframes. The assessed providers employ more than 170,000 mainframe programming experts, including 60,000 COBOL programmers. In operations, more than 53,000 experts keep mainframes running. They have five years of experience on average. Only 6 percent of the mainframe operators have more than 14 years of experience. However, these numbers need to be put in perspective. COBOL skills are just as rare as SAP. A LinkedIn search returns more than 350,000 people with COBOL skills, 288,000 experts in ABAP (SAP programing language), 3 million C# programmers, and an astonishing 10 million people with Java skills.

Offshoring is a solution for skills shortage: All participating service providers have global operations and COBOL delivery capacity in India. The assumption that a COBOL career is not interesting to youth generations is valid in the U.S.; however, global companies have found ways to attract and retain talents to work on mainframes and COBOL.

Knowledge versus innovation dilemma: Knowledge retention can be a challenge for clients that migrate from COBOL to Java because newly hired programmers do not understand the business and the company may not have a career path to offer to experienced COBOL programmers. A few solutions that compile COBOL to Java enable the co-existence of both programmers for a smooth transition, enabling for knowledge transfer.

The future of COBOL is Java: Java is by far the preferred destination language when moving off the mainframe. Other languages include .NET, C#, Python, and Powershell (the last two for scripting Batch jobs.). Code re-engineering and code conversion tools provide automation to replace COBOL with Java. It performs well on any cloud and any relational database. Application development tools can handle both languages, providing a smooth transition for application development shops.

Estimating project cost: Vendor and providers usually mention lines of code (LOC) as the base for cost estimation (76 percent of the respondents). However, complexity, tools, and size have an additional impact on pricing. Some statistics include:

• Modernization and code refactoring cost: US$0.25 to US$2.30 per LOC; project duration: 2 to 36 months; and project cost: US$100,000 to US$25 million.
• Transformation and code conversion cost: US$0.50 to US$8.00 per LOC; project duration: 6 to 60 months; and project cost: US$100,000 to US$50 million.

Estimating project viability: Mainframe MIPS measures hardware capacity; it is not used for project estimations. However, it provides the first cost estimate for the cloud. A rough estimation is one x86 core in the cloud can replace 50 to 100 MIPS mainframe.

• Top 100 mainframe clients manage more than 50,000 MIPS, with few of them operating more than 200,000 MIPS.
• Very large MIPS client manages 10,000 to 50,000 MIPS.
• Large clients manage 5,000 to 10,000 MIPS.
• Mid-sized clients manage 2,000 to 5,000 MIPS
• Small clients manage less than 2,000 MIPS.
• Top mainframe outsourcing providers manage more than 300,000 MIPS each, up to millions of MIPS.
• The average outsourced mainframe has 4500 MIPS per client.
• The U.S. concentrates 60 percent of global mainframe MIPS.

Clients running less than 5,000 MIPS should consider migrating their mainframes to the cloud. Any of the migration options are viable and cost-effective, providing short-term ROI. Clients hosting 10,000 MIPS and more can consider MFaaS as the first move for cost saving while assessing the modernization and transformation options. Top 100 mainframe clients run mainframe farms, not single monoliths. Outsourcing is a good option to reduce cost, while offshoring eliminates the skill shortage risk. Simultaneously, top mainframe clients can consider clustering their mainframe systems around similar business functions to study each cluster separately.

Financing mainframe modernization is a challenge: Many companies consider mainframe modernization a low return investment. A CIO of a large bank with more than 100,000 MIPS responded to our question saying, “my mainframe is running the bank’s support functions, it is certainly not a problem, and I am not going to invest in it, not even to turn it off.” Vendors of modernization tools responded that, in most cases, the CIO sees the value but considers that the risk is an impediment. Vendors and service providers are working on making the projects faster, secure, and cost-effective to enable mainframe modernizations. 

Consider self-financing the modernization: Some providers of application management and support services (AMS) propose deals that include mainframe modernization in the AMS goals. As a result, clients take the maintenance budget they already have and use it to move applications to low-cost platforms and code  that is easier to maintain. Their threeyear deal becomes a transformation program.

Access to the full report requires a subscription to ISG Research. Please contact us for subscription inquiries.