Kyndryl will use Google’s AI tools in mainframe modernization projects

Kyndryl is partnering with Google to help joint clients modernize their mainframe workloads.
The collaboration builds on an existing partnership that launched after Kyndryl became an independent company in 2021. Up to that point, the consultancy functioned as IBM’s IT infrastructure services unit. Since the spinout, Kyndryl has boosted the number of Google Cloud certifications in its workforce to more than 6,000 and launched an AI consulting collaboration with the search giant.
The companies’ partnership expansion extends their AI work to the mainframe ecosystem. Mainframes, which are made almost exclusively by former Kyndryl parent IBM, continue to power many important applications. One estimate puts the number of credit card transactions that are processed with the help of such machines at 90%.
One reason mainframes continue to find use is that they’re highly reliable. IBM’s latest z16 mainframe is advertised as providing nine nines of availability, which means it experiences about three milliseconds of downtime per year on average. But operating such specialized hardware also comes with challenges, which is why many companies are opting to move their mainframe workloads to the cloud. That’s the task Kyndryl plans to ease through its newly expanded partnership with Google Cloud.
As part of the collaboration, the former IBM unit is launching a new offering called the Mainframe Modernization with Gen AI Accelerator Program. It will help companies move applications and data from mainframes to Google Cloud. In conjunction with the offering’s introduction, Kyndryl announced that it has been certified as a specialized Google Cloud partner for Gemini models. The consultancy plans to use the search giant’s AI technology to speed up the mainframe-related projects it takes on for clients.
Like most IT consulting engagements, a mainframe modernization program usually starts with a review of the client’s objectives and technology stack. “Before proceeding with a modernization project, we want to help define what is the best course of action for a customer, based on their specific goals and needs,” a Kyndryl spokesperson told Boardroom Insight. “Therefore, we will start with an examination of the customer’s mainframe data and applications with the aim to assess what a mainframe modernization project would look like, and then build a blueprint and plan for that journey.”
Kyndryl’s consultants will carry out such assessments with the help of a Google Cloud product called the Mainframe Assessment Tool, or MAT. It uses AI to collect technical information about a mainframe application’s code base. MAT can determine how many lines of code the application includes, what programming languages the code is written in and the cyclomatic complexity of the software. That kind of information makes it easier for consultants to figure out details such as the amount of time it will take to rewrite a mainframe application.
Besides planning mainframe modernization initiatives, Kyndryl can also carry out the hands-on technical work involved in the process. Its consultants will use AI to expedite that work, the Kyndryl spokesperson explained.
MAT “uses generative AI to analyze a mainframe application and codebase, and create an output that helps Kyndryl determine the best path for modernization,” the spokesperson said. “For example, if the best path is to refactor the application, Kyndryl might use Google’s G4 solution; if the best approach is to replatform, Rocket Enterprise Server might be used; and either of these solutions might include Dual Run.”
Replatforming is a relatively limited software overhaul that involves shifting an application to new infrastructure without extensively changing its code base. Refactoring, in turn, is the process of rewriting a workload from the ground up. The latter approach is more complicated but can enable ported applications to run more efficiently on the new infrastructure.
“If MAT’s output suggests a path to rewrite, then Kyndryl would take the documentation produced by MAT and rewrite the application,” the company explained.
G4, Rocket Enterprise Server and Dual Run, the three products that the consultancy will use during such client engagements, each focus on a different set of technical tasks.
G4 is an AI tool that Google obtained through 2020 acquisition. It can switch code written in mainframe-specific languages such as COBOL to Java, which speeds up refactoring. Rocket Enterprise Server, in turn, is mainly geared towards replatforming projects. It provides an abstraction layer that allows mainframe applications to run on modern infrastructure.
Dual Run, the third product Kyndryl will use, makes it easier to test the reliability of mainframe workloads ported to the cloud. Using it requires setting up two separate copies of an application: one version that runs in the cloud and another that runs on a mainframe. Dual Run searches for bugs by comparing the two copies’ behavior. If the cloud-based version of an application generates the same output as the original mainframe version in response to user requests, developers can draw the conclusion that the former workload is functioning reliably.
Kyndryl says that it has already worked with Google to help a joint customer, an insurer, move a set of mainframe applications to the search giant’s Google Distributed Cloud platform. The project saw the consultant’s professionals switch the workloads’ code base from COBOL to Java.