Infrastructure Modernisation: A Year-Long Journey of Migrating Research Projects and Core Services

Introduction

Over the past year, our team has been working on a project to migrate our research projects from a private VMware infrastructure to a centrally managed OpenStack infrastructure within King's College London's e-Research department. This migration represents not just a change in hosting, but a complete modernisation of our digital estate, encompassing projects spanning over two decades of digital humanities research.

To date, we have successfully migrated the majority of our approximately 85 projects, with only four remaining. For these remaining projects, we have implemented static copies to ensure continued access to their public interfaces, while we complete their full migration, albeit with limited functionality.

Project Scope and Challenges

The scope of this migration was particularly complex due to the diversity of our digital estate:

• Technologies ranged from Java-based webapps to Django applications and WordPress sites
• Project ages varied significantly, with some dating back more than 20 years
• Version control systems included SVN and Mercurial repositories that needed conversion to Git
• Legacy systems requiring containerisation while maintaining functionality

Background and Context

Our infrastructure and methodological approaches have been documented in several publications. For readers interested in a deeper understanding, we recommend:

• Ciula and Smithies, “Sustainability and Modelling at King’s Digital Lab: Between Tradition and Innovation”, in: Nyhan, Rockwell, Sinclair (eds.): On Making in the Digital Humanities. Essays on the Scholarship of Digital Humanities Development in Honour of John Bradley. London: University College London Press, 2023. https://doi.org/10.14324/111.9781800084209
• Smithies and Ciula. Humans in the Loop: Epistemology & Method in King’s Digital Lab, Schuster & Dunn eds. Routledge International handbook of research methods in digital humanities: 155-172, 2020. https://doi.org/10.4324/9780429777028-13
• DigiSpec: Scoping Future Born-digital Data Services for the Arts and Humanities: Case Reports. Zenodo, 31 Aug. 2022, doi:10.5281/zenodo.4716148.

Modernisation Strategy

Containerisation and Standardisation

A key aspect of our modernisation effort was the comprehensive containerisation of all active projects. This process revealed several challenges:

• Finding compatible Docker images for legacy software without extensive codebase modifications
• Addressing platform-specific issues, particularly with Apple Silicon machines
• Managing dependency updates across diverse technology stacks
• Dealing with encoding issues that manifested differently in OpenStack/Ubuntu compared to local environments

Infrastructure as Code

We significantly improved our deployment process by implementing infrastructure as code practices:

• The full infrastructure is now described in Terraform code, including the definition of all services, networks and security rules in conjunction with Ansible to manage the virtual machine itself. In addition, this ensures that VMs are auto-patched, typically weekly, contributing towards security
• Developed reusable GitLab CI templates for common deployment patterns
• Created standardised Docker templates for common service types (Django, Tomcat, Cantaloupe image server, nginx, proxies)
• Simplified deployment to a git push operation
• Implemented consistent environment management across all projects

Technical Insights and Lessons Learned

Throughout the migration, we encountered and resolved various technical challenges:

Platform-Specific Issues

One of the most interesting aspects of the migration was dealing with platform-specific behaviours that only manifested in the OpenStack environment. A particularly notable example was with Kiln/EFES projects, where XSLT processing would fail with non-obvious errors. The solution required deep diving into Cocoon's pipeline architecture and modifying how templates were processed. We also met with some architectural changes regarding presenting web sites to the Internet. We no longer place project servers directly in the Internet, but route all public web access via web proxy servers.

Java Application Challenges

Java applications presented unique challenges, particularly around class-loading behaviour. Loading order differed between local development and OpenStack environments, requiring careful management of JAR file placement to ensure consistent behaviour across environments.

Docker and Container Orchestration

Container-related issues required careful attention to platform compatibility, environment synchronisation, and dependency management. This was particularly challenging with legacy applications that required specific versions of dependencies or had complex runtime requirements.

Static Site Conversion: A Sustainability Strategy

The Decision to Go Static

Our approach to static conversion emerged as a key sustainability strategy, driven by practical considerations:

• Security vulnerabilities in legacy systems
• Limited resources for ongoing maintenance
• End of project funding cycles
• Technical feasibility issues
• Cost of maintaining dynamic infrastructure

If interested, you can find out more about our static-first development approach.

Technical Implementation Process

1. Initial Site Capture

   • Created complete site snapshots as the first step
   • Documented existing functionality and user flows
   • Identified critical features that needed preservation
   • Used tools like Sitesucker for creating comprehensive static snapshots

2. Technical Challenges

   • Sites with AJAX-dependent interfaces required special handling
   • For complex dynamic functionality:
     • Captured all necessary data states
     • Reimplemented critical features using client-side tools
     • Implemented modern static search functionality using pagefind where needed
     • Simplified interfaces where appropriate
   • Preserved core user experience while accepting some functionality trade-offs

3. Sustainability Benefits

   • Dramatically reduced security attack surface
   • Minimal ongoing maintenance requirements
   • Lower hosting costs
   • Improved performance
   • Extended project lifespan

Stakeholder Engagement

The conversion process involved careful stakeholder management:

• Lab support team led communications with project stakeholders and senior management in our Faculty as well as relevant IT departments
• Clearly explained technical implications and benefits
• Focused on long-term preservation benefits
• Generally received positive reception due to extended project lifespan
• Transparent about functionality changes or limitations

Impact and Benefits

The modernisation effort has yielded several key benefits:

• Improved maintainability through standardised deployment processes
• Enhanced sustainability through containerisation
• Reduced operational costs by optimising resource allocation
• Better documentation and reproducible infrastructure
• Simplified ongoing maintenance and updates

These benefits allows us to pass on cost savings to our partners and Faculty, for whom we maintain these projects as well as align with central e-Research facility and services.

Team and Timeline

The project, so far, has presented significant resource challenges:

• Team composition shifted during the year-long project
• Balancing migration work with ongoing active projects
• Maintaining service continuity while implementing changes
• Managing knowledge transfer during team transitions

Looking Forward

This migration has positioned us better for future sustainability:

• Standardised processes for new project onboarding
• Improved infrastructure flexibility
• Better prepared for future technological changes
• More resilient and maintainable digital estate

Conclusion

While more time-intensive than initially anticipated, this migration project has positively transformed our infrastructure and processes. The standardisation and modernisation efforts have not only improved our current operations but have also established a robust foundation for future digital humanities projects at KDL.

Our shift towards a static-first approach, combined with comprehensive containerisation and automation, ensures that our digital research outputs remain accessible and maintainable for years to come. As we work to complete the migration of the final projects, the successful transformation of the majority of our digital estate demonstrates the value of systematic modernisation in research infrastructure (next major upgrade expected in 3 years time) and good team work!