How a Fortune 500 retailer worked with Mechanical Orchard to safely and quickly move from the mainframe to the cloud
A Fortune 500 retailer wanted to move from the mainframe to the cloud to reduce costs, increase agility, and unlock growth and innovation opportunities.
A consultancy-led “lift and shift” attempt ultimately failed at great cost and wasted time. Mechanical Orchard was then entrusted to deliver the migration via an AI-enhanced, iterative and reverse-engineered approach which instead addressed risk continuously throughout the process.
The Company is successfully offloading and accelerating a number of mainframe workloads into modern software on the cloud, with no disruption to the business.
From its origins as a corner grocery store in the 1920s, the Company thrived through the decades, becoming one of the most recognizable retail brands in the United States.
At the heart of the Company’s operations was the mainframe estate, which supports core capabilities such as supply chain, merchandising, finance, enterprise, and store management. Tens of millions of dollars are spent annually on maintaining this infrastructure. Critically, the mainframe is maintained by nine engineers, several of whom are approaching retirement. These individuals possess unique context and lived experience of how the mainframe system evolved over the decades.
This was not their first effort. Most recently, the Company engaged a global systems integrator to take the industry-standard approach: transpile COBOL[1] to Java.
At the end of three years and $40 million, the transpiled code ran at a tenth of the performance of the mainframe, rendering the code useless. No functioning software ever entered the production environment.
With an increasingly urgent need to deliver promises made to investors, the IT team was keen to avoid the “COBOL in the cloud” — or “JOBOL” — approach, and even more adamant to avoid the “digital transformation” efforts that had resulted in tens of millions of wasted dollars. As a result, they were receptive to Mechanical Orchard’s (MO) strategy, which promised an incremental and accelerated approach to modernizing systems that provided ongoing proof of feasibility, with no disruption to day-to-day operations.
About the Mainframe
The Company’s mainframe system is a z/OS[2] and COBOL-based system of applications and databases that had grown close to seven million lines of code by 2023.
Comprising the mainframe landscape are thousands of batch processes and CICS[3] screens integrating with internal and external dependencies, orchestrated by D-Series and backed by VSAM[4] and DB2[5] data stores. Altogether, the mainframe system consumes roughly 6,000 MIPS[6] across 5 LPARs[7].
Mechanical Orchard’s incremental rewrite modernization approach minimizes risks and disruption through three principles:
Typically, the amount of risk increases at a faster pace than the number of workloads addressed, thanks to compounded risk.
A better way is to get each individual workload all the way into production. While the first workload can take a few months to be ready for cutover, subsequent workloads cutover at an accelerating pace.
The process of creating a behavioral twin begins with establishing a parallel cloud environment in which to run each incremental, modernized workload. Beginning in March 2023, the MO team worked closely with the Company to establish a VPN connection between the target cloud environment and the mainframe to create an automated DevSecOps continuous delivery pipeline, an automated test environment, and an end-to-end path to production.
Next, the Company collaborated with MO to identify a place to start based on the mass of the system (its size) and the complexity of the system (its interdependencies), both of which contribute to overall risk. Based on the most recent failed experience, the Company felt that addressing complexity first would be most beneficial.
Together, the Company and Mechanical Orchard zeroed in on the physical inventory management system identified via a detailed analysis and visualization of the mainframe system, created through Mechanical Orchard’s LLM-assisted technology. From this system, the teams jointly prioritized which workloads to begin with.
At the top of the list was a high-performance batch process: the same one that precipitated the failure of the last modernization attempt.
This pan-system view also helped prioritize the other workloads within the mainframe, which meant that additional workload migrations could begin in parallel.
Selection made, the Mechanical Orchard team began learning as much about the batch process as possible through data ingestion and source ingestion. With data ingestion, MO identified the databases the process reads and writes to, combed through documentation, and interviewed users. With source ingestion, MO examined the COBOL code, running it through specialized, LLM-based tooling to better understand its functionality. While this process typically can take several months through the use of existing tools, MO is able to gain a comprehensive understanding in as little as a few days.
This created a very detailed set of specifications with which Mechanical Orchard could build tests to ensure a faithful reproduction of the batch process.
Once the batch process was well-understood, MO created a behavioral replica in a secure cloud environment, where it was subjected to rigorous automated equivalence testing against live production data, side by side with its legacy counterpart, before cutting over to the target cloud environment.
By the end of April, the replicated job was running
at 100% equivalence against full production runs.
In the meantime, work on subsequent batch processes had begun.
By mid-June, the MO team was working on the four riskiest workloads in the application: the first workload officially cutover from the mainframe on June 27, 2023.
In less than four months from the start of the overall engagement, the Company could see that this approach was indeed working: modern software was in production, working in concert with the existing system.
And because their riskiest workload was now in production, the Company had confidence that they had solved the hardest problems first: everything ahead would be a far more straightforward exercise of accelerating the migration of workloads into the new production environment.
Above all, the IT team at the Company trusted that the entire mainframe offload could be completed safely and reliably, resulting in a truly modernized IT estate that can easily respond and adapt to rapid shifts in the market — all at a fraction of the time of traditional approaches.
Chief Enterprise Architect
While the first batch process took four months to reach production, subsequent mainframe jobs are now being delivered at a pace of two per week— from initial comprehension to fully functional software running in production. Work on the remaining components, including CICS screens and data migration, began in September 2023 and has continued to accelerate.
By September 2024, we had successfully modernized and retired one of their largest, most critical global inventory and supply chain systems— responsible for vendor replenishment—migrating it to the cloud with zero disruption and industry- leading SLA performance.
Batch job cycle time has improved by 265%, and throughput has increased by 347%.
With the introduction of the Imogen, the Mechanical Orchard mainframe modernization platform, we anticipate that two of the largest remaining mainframe applications are on track to be fully modernized by January 2026—roughly 65% faster than traditional methods, all while avoiding business disruption and maintaining operational continuity.
Figure 5:
Velocity increases exponentially with each subsequent workload.
As organizations grow, silos can appear, making it difficult for IT teams to feel like they are a core part of strategic, exciting projects. This issue, along with the Company’s over-reliance on mainframes, was identified as a blocker to the successful recruitment of talented technical staff.
One effect of the Mechanical Orchard approach is the team’s renewed confidence and enthusiasm for working on a high-profile project that had previously been viewed as a hopeless cause.
This palpable energy is fueled by:
Although calculations are preliminary, the reduction in lines of code is substantial, between 50% and 65% so far, with some batch jobs seeing a reduction of up to 80%.
Less code means reduced complexity. This is particularly helpful with promoting large system stability, improving and simplifying maintenance, and, ultimately, performance. In addition, fewer lines of better quality, clean and optimized code helps eliminate technical debt, decrease the threat surface area for cyberattacks, and most importantly, makes it far easier and faster to pivot the system to adapt to business needs.
Moreover, modern languages are more powerful and expressive; similarly, workloads rewritten in modern languages are far more malleable, understandable, and exciting to work with.
This case study highlights how Mechanical Orchard is helping the Company successfully transition from its legacy mainframe system to a modern, cloud- based infrastructure.
While past efforts — including significant investments in time and money — have taken years to determine whether the transition is working or not, the MO approach demonstrates genuine progress continuously, with no disruption to daily operations.
This method has not only improved system performance and reliability but has also restored confidence and morale within the IT team and wider organization that the Company is set up for the future.
We believe that every company deserves to realize their vision, free of constraints from the past. Our team's approach, technology, and experience can help them move into this evolving version safely, reliably, fearlessly.
About the research
• Research conducted by TechTarget’s Enterprise Strategy Group
• Survey of 150 Senior IT decision makers (C-suite to directors) across large enterprise (2,000+ employees) organizations with>$1 billion total annual revenue
• North America (US, Canada, 67%), Western Europe (UK, 33%)
• Multiple industry verticals across the public and private sector
• Field dates: 4/2/2024 – 4/24/2024
• 82% of respondents had undergone a legacy modernization project in the last 12 to 24 months