AWS Offers Mainframe Migration Service. Is IBM i Next?
December 6, 2021 Alex Woodie
Amazon Web Services, the world’s largest cloud service provider, last week unveiled a new service to help IBM mainframe customers modernize their workloads and move them to the cloud. Will the cloud giant, which says it’s completely driven by the needs of its customers, launch a similar service for midrange workloads running on IBM i?
AWS Mainframe Modernization is a new service AWS launched at its re:Invent conference last week that aims at making it “faster and easier to migrate mainframe and legacy workload to the cloud.” In addition to AWS components, the service includes roles for systems integrators, such as Accenture and DXC Technology, who will provide much of the grunt work in moving these complex applications into their cloudy new homes.
The service includes a runtime environment on EC2, with certain configurations of compute, memory, and storage, to run both the refactored and re-platformed applications. It also brings all the cloud bells and whistles (provisioning, load balancing, auto-scaling, etc.) and security and monitoring that such critical enterprise applications demand.
According to AWS, the mainframe modernization service gives users the option of “refactoring” their mainframe workloads by transforming them into Java-based cloud services, or “re-platforming” them with minimal code changes. The former is a more involved process that involves breaking down the legacy code and rewriting it to run efficiently in modern container-based environments, while the latter is a quicker “lift and shift” type of approach.
Adam Selipsy, the former head of Tableau Software who took over the CEO job at AWS earlier this year after Andy Jassy took Jeff Bezos’ place as the head of the world’s fifth-largest company, said AWS is seeking to make the cloud cost effective “for every workload,” and that includes IBM System z workloads.
“Mainframes are expensive. They’re complicated. And there are fewer and fewer people who are learning to program COBOL these days,” Selipsky said during his keynote at re:Invent last week. “This is why many of our customers are trying to get off of their mainframes as fast as they possibly can to gain the agility and the elasticity of the cloud.”
AWS has already worked with its partners to migrate mainframe applications to cloud, and has several case studies on its website that discuss the projects in some depth.
For example, AWS helped the New York Times migrate a COBOL-based CICS and z/OS application that was handling billing, invoicing, customer accounts, delivery routing, product catalog, pricing, and financial reporting to the AWS cloud. The application was composed of 2 million lines of COBOL, used a BMS-based 3270 interface, stored data in the VSAM KSDS format, and had 600 batch jobs, which were handled via JCL and CA7, according to the case study on AWS’s website.
The project was split into two phases, including the automated refactoring, which began in 2015 and took two years, and the AWS migration, which started in 2017 and lasted eight months. All told, testing accounted for approximately 70 percent to 80 percent of the time spent on the project, AWS said. The new Java-based app, called Aristo, went live in August 2017. It has reduced the newspaper’s IT costs by 70 percent, AWS claims.
AWS also has case studies detailing mainframe migrations completed with the United States Air Force and Kmart, which it completed with partner Micro Focus. Clearly, the cloud giant and its partners have some experience with these migrations, which can be painful, messy, and time-consuming, and which often fail, as the New York Times’ first migration in 2006 did. The fact that AWS is rolling out a service indicates the company believes it has settled on an approach and methodology that it thinks can be successful among a broader set of mainframe shops.
“The AWS Mainframe Modernization can cut the time that it takes to move mainframe workloads to the cloud cut by up to two-thirds, thanks to a complete set of development, test, and deployment tools, and also mainframe compatible runtime environments,” Selipsky said. “Mainframe modernization helps you assess, helps you analyze your mainframe application for readiness, choose the path you want – re-platform or refactor – and then come up with a plan.”
Will AWS offer up a similar program for the mainframe’s little brother, the IBM i server? At this point, there’s no indication that it will. However, if the company truly just delivers what its customers ask it to deliver, it would seem to be a possibility for the future.
While there are fewer IBM mainframes running than IBM i servers and predecessor machines, they typically have bigger workloads. Mainframes are also heavily used in financial services and healthcare industries, and the companies tend to be larger than the IBM i strongholds in manufacturing, distribution, and retail.
With that said, there are plenty of big IBM i deployments in the world that are facing very similar circumstances as their mainframe counterparts. They’re running proprietary platforms with applications written in older languages that aren’t being heavily taught in school. IBM has embraced open source technologies in an attempt to bolster the capabilities of the machines’ and the abilities of individuals in the community, with mixed success.
AWS’s competitors have danced with the midrange server to some extent. Back in 2019, Microsoft – which has a complicated history with the IBM midrange server – unveiled the latest program to attract IBM i shops into its arms. It didn’t look much different than what AWS is now doing with the mainframe: involve partners to do the heavy lifting of refactoring code into a newer language (.NET languages, but also Java and PHP) and more “modern” middleware, while hosting the final migrated applications in Azure. Microsoft is also working with Seattle-area partner Skytap to bring IBM Power processors (and native IBM i applications) into Azure, where it can be scaled up and down and managed right alongside X86-based Azure apps.
Meanwhile, Google Cloud appear to have hit a snag with its plan to spin up IBM i environments in its cloud. It’s already running Power processors, but it’s limited to running AIX and Linux; the IBM i environment is stuck in alpha.
AWS has become a giant company, with 1.5 million customers and a $68 billion run rate. It offers more than 200 pre-built managed services, ranging from databases and data lakes to analytics offerings, and more than 400 runtime instances on EC2. It runs a multitude of processors besides X86, including ARM, AMD, Nvidia, and even Mac, and has even started developing its own custom silicon that it says can run customer workloads even more efficiently.
The company relentlessly maintains that it only builds what its customers ask it to build; there is no ”build it and they will come.” With calls for more high-level services instead of building blocks, its CTO, Werner Vogel, blamed customers. “It’s your fault” there are 200 Lego building blocks, he said this week at re:Invent.
Could there be room for an IBM i service at AWS? It doesn’t appear to be in the cards at the moment. But if AWS customers who also have a smattering of IBM i applications demand it, it’s conceivable that AWS and its partners could take notice.