Swimm Tackles Legacy Mainframe Modernization with AI-Powered Documentation

Key Concepts

Mainframe modernization, static analysis, AI (LLMs), code documentation, business logic extraction, legacy systems, COBOL, PL1, assembly, autosync technology, CI/CD pipeline integration, risk reduction, modernization project failure, code coverage, efficiency gains.

Mainframe Modernization Challenges

• Black Box Problem: Mainframe code is often a "black box" due to the retirement of original developers and lack of understanding of the underlying business logic.
• Documentation Gaps: Existing documentation is often outdated or incomplete, especially after incidents like the World Trade Center bombing, where documentation was lost.
• Skill Shortage: There's a dwindling pool of COBOL and other legacy language developers, making it difficult to maintain and modernize mainframe systems.
• High Failure Rate: Modernization projects involving mainframes have a high failure rate (reportedly 79%) due to the inability to fully understand the business logic.
• Fear of Change: Companies are often hesitant to modify mainframe systems due to the fear of unintended consequences.

Swim's Solution: Static Analysis and AI-Powered Documentation

• Combination Approach: Swim combines static analysis with AI (specifically LLMs) to understand and document mainframe code.
• Static Analysis First: Static analysis is used to analyze the entire codebase, structure it, and create a scaffold for documentation. This is crucial because LLMs are not well-trained on legacy languages.
• LLM for Natural Language Translation: LLMs are then used to translate the structured data into human-readable documentation, making it accessible to developers and business analysts.
• Business Logic Extraction: The goal is to uncover the business logic and business rules embedded within the mainframe code.
• Customization: Swim customizes its solution to meet the specific needs of each client, including their code frameworks and documentation templates.

Autosync Technology for Dynamic Documentation

• Code-Document Connection: Documentation is directly linked to the code, including code snippets and descriptions.
• Proprietary Autosync: Swim's autosync technology automatically updates documentation whenever the code changes.
• No Manual Intervention: This eliminates the need for manual updates and ensures that the documentation always reflects the current state of the codebase.
• IDE and CI/CD Integration: Autosync can be integrated into a developer's IDE or the CI/CD pipeline.
• Commit-Based Tracking: The CLI tool tracks code commits and identifies if a commit has affected documentation.
• Pull Request Blocking: The system can even block pull requests to enforce documentation updates.

AI Technology Details

• Standard LLMs: Swim uses standard LLMs for natural language translation.
• Customer-Hosted LLMs: The solution can connect with customer-hosted LLMs (e.g., Azure OpenAI) for sensitive code.
• Local Static Analysis: Static analysis runs locally on a developer's machine.
• Secure Loop: The entire process happens within the customer's network for security.

Modernization Paths and Benefits

• Understanding Business Logic: Even if companies choose to keep the code on the mainframe, understanding the business logic is crucial due to the shortage of COBOL developers.
• Cloud Migration: The extracted business logic can be used to migrate code to the cloud.
• Language Translation: The code can be translated into more modern languages.
• Risk Reduction: Understanding the business logic reduces the risk associated with modernization projects.
• Efficiency Gains: Automating documentation significantly reduces the time required to understand and modify mainframe code.

Project Status and Metrics

• Active Pilots: Swim has active pilot projects running with customers.
• Customization Phase: The initial six-week phase involves customization to the client's specific needs.
• Code Coverage: The solution has been successfully tested on codebases up to 10 million lines of code.
• Efficiency Improvement: Documenting a single program manually can take two to three weeks, while Swim can do it in minutes.

Swim's Background and Evolution

• Documentation Focus: Swim has always been a documentation solution.
• AI Integration: AI elements have been added to enhance the documentation process.
• Backfilling Documentation: Swim focuses on backfilling documentation gaps for large existing codebases where business logic has been lost.
• Developer-Centric Approach: The technology is designed to help developers create, maintain, and find documentation more easily.

Conclusion

Swim offers a solution to the challenges of mainframe modernization by combining static analysis and AI to extract and document business logic. Their autosync technology ensures that documentation remains up-to-date with code changes, reducing risk and improving efficiency. While still relatively new to the mainframe space, Swim's approach shows promise in helping companies understand and modernize their legacy systems.