# Experiments in sustainable digital humanities projects January 27, 2026 From January to April 2025, Digital Preservation Services partnered with Harvard University IT (HUIT) Arts and Humanities Research Computing on a pilot to reduce the resource load for maintaining Harvard’s digital humanities projects. The four-month sprint was generously funded through the HUIT Emerging Technology and Innovation Program ([ETIP](https://www.huit.harvard.edu/etip)), and the team produced a sample set of statically maintained digital humanities projects and a guide to [Designing Sustainable Digital Projects](https://digitalhumanities.fas.harvard.edu/resources/designing-sustainable-projects/). Read on to learn more about the project -- or even better, come talk about it in person at the [**Love Data Rodeo**](https://libcal.library.harvard.edu/event/16237169) **in Lamont Library Forum Room on February 12th from 12-2pm during Love Data Week**. ## What prompted this project? In late 2025, HUIT’s Digital Arts and Humanities Research Computing (DARTH) team consulted with Digital Preservation Services on a challenge: how to preserve custom digital projects after development, without the long tail of maintenance resources that accrue over time. Harvard faculty work with DARTH to build innovative digital scholarship tools, both as research outputs and as learning tools in the classroom. These projects often involve focused collaboration and investment up front, with principal investigators, postdocs, research assistants, and software engineers that span Harvard boundaries. But once they are launched and no longer in active development, preserving these digital outputs is more complex and costly than preserving traditional print-based scholarship. There are ongoing expenses of cloud hosting, mounting security risks that need to be curbed as projects age, and the vulnerability of shifting technology stack dependencies and custom code. As projects accrue, the DARTH team balances the substantial operational workload of maintaining them against supporting new project development. > Harvard’s digital scholarship preservation landscape includes numerous options for preserving individual project assets – such as datasets, scripts, still images – however, there are few options to preserve the interactivity of full-stack scholarly applications over time. Arts and Humanities Research Computing have experimented with flattening sites and removing database compute layers, but this is a heavy-handed and bespoke solution that often results in a loss of functionality. Harvard’s digital scholarship preservation landscape includes numerous options for preserving individual project assets – such as datasets, scripts, still images – however, there are few options to preserve the interactivity of full-stack scholarly applications over time. ## Pilot overview The HUIT ETIP program supports short-sprint pilots, and we planned our four-month timeline around experimenting with solutions to the question: **how can we minimize the long tail of maintenance costs, uncertain futures, and unintended loss for digital scholarship projects?** The intended outcomes of the project were not constricted to improving Harvard’s ability to sustain projects; rather, we wanted to focus on using project resources in a sustainable way. To this end, we sought practical solutions to sustain static projects and reduce maintenance costs, as well as a framework for when to sunset, sustain, or preserve a project, based on the project’s objectives, resources, and audience. ### Pilot components The primary strategies we investigated were: - A **sustainability framework** to guide faculty and technical staff in evaluating different sunsetting and preservation options. The recurring refrain in digital humanities over the years has been “plan for sunsetting or sustaining intentions at the project outset.” Faculty are often unaware of the total costs of stewardship associated with maintaining projects and do not account for those recurring costs after the active project phase. Few people have a landscape view of the various sunsetting or maintenance options, as well as which variables influence successfully selecting one of those options. - **Web recording platforms** that could act as both a low-effort solution for sustaining static sites and be a preservation solution if they are retained for long-term. [Archive-It](https://archive-it.org/) and [Webrecorder](https://webrecorder.net/) were evaluated based on the following criteria: cost, searchability, flexibility, potential integration into existing Harvard systems, automation, performance, and fidelity across different projects. - **Ephemeral environments** as a method of reducing the total cost of project ownership for complex, full-stack applications. An ephemeral environment is a short-lived, isolated deployment of an application, which was an appealing solution because finished research projects tend to be containerized, the data is often public and can be pulled into a static element like a csv, and they are low-traffic and standalone (no APIs). This was the most experimental component of the project, and the DARTH team ended up isolating the tests to [Google Cloud Run](https://cloud.google.com/run?utm_source=google&utm_medium=cpc&utm_campaign=Cloud-SS-DR-GCP-1713658-GCP-DR-NA-US-en-Google-BKWS-BRO-CloudRun&utm_content=c-Hybrid+%7C+BKWS+-+BRO+%7C+Txt-AppMod-Serverless-Cloud+Run-765571308157&utm_term=google%20cloud%20run&gclsrc=aw.ds&gad_source=1&gad_campaignid=22980675520&gclid=CjwKCAiA7LzLBhAgEiwAjMWzCGkk8CApWk-ffWwKULHqmw9TskvAG1M3_ldHT6tbKBxIxZeXJ4yPrBoCXmcQAvD_BwE) experiments using a “scale-to-zero" approach of automatically reducing application resources when use is inactive and rapidly scaling back up when traffic is active. ### Participants The core team consisted of Arthur Barrett (Academic Technology for the Faculty of Arts and Sciences), Cole Crawford (Arts and Humanities Research Computing), and Tricia Patterson (Digital Preservation Services). Consultants brought in for their expertise and feedback also spanned HUIT and Harvard Library: Bill Barthelmy (Academic Technology), Matt Cook (Digital Scholarship), Jeremy Guillette (Academic Technology), and Chelcie Juliet Rowell (UX and Discovery). Beyond that, several other Library colleagues offered valuable input, such as Skip Kendall and Juliana Kuipers in Harvard University Archives and Vanessa Venti and Anna Van Someren in UX and Discovery. The team selected multiple scholarly projects that were suitable for testing different experimental components, and we benefited from the widespread scholarly collaborations as well. For example, Micha Broadnax (Gutman Library) and Tracy Stuber (Academic Technology), who were in the planning phase of a new custom project, tested the sustainability guidance. ## Outcomes and recommendations ### Designing Digital Projects Guide There has been a lot of excellent progress towards sunsetting or preserving elements of digital scholarly projects, such as the [Socio-Technical Sustainability Roadmap](https://sites.haa.pitt.edu/sustainabilityroadmap/) and [The Endings Project](https://endings.uvic.ca/). These resources offer strong general frameworks, so we drew on this this existing work to create guidance about specific options available to Harvard faculty, which resulted in the [**Designing Sustainable Projects**](https://digitalhumanities.fas.harvard.edu/resources/designing-sustainable-projects/) **guidance, now hosted directly on the DARTH website**. The guidance includes: - A primer on risks and challenges for sustaining digital scholarly projects. - A menu of sunsetting and sustainability options to help project teams target what “end of life” can look like for the project, as well as consider the socio-technical factors to incorporate into the project design to achieve their intended outcome. - A Digital Sustainability Dossier to guide discussion and documentation of everything necessary to achieve the “end of life” target and to empower potential future stewards of the project to successfully meet targeted outcomes. - Citations and further reading, for those that want to learn more. Two project teams piloted the guidance, and we received feedback that it promoted visibility of costs throughout the project lifecycle, set realistic expectations about technical decisions, and empowered teams to discuss options like sunsetting or static maintenance. ### Web archiving for maintenance and preservation Web archiving can be an effective method of maintaining interactivity of less complex scholarly projects which are relatively static, meaning they are not generating new outcomes or relying on continuously updated external databases. Both Archive-It and Webrecorder, two popular tools in the archiving community, are useful for maintaining and/or preserving different kinds of projects, depending on long-term needs. ***Original site, before needed maintenance*** Below is an example of a Harvard site from 2013-2014, Hearing Modernity, that we tested against both web archiving tools. The first image is the site prior to undergoing some routine maintenance, and it demonstrates the continuous care required for projects after they are out of the active development phase. Domains can expire, styling errors accrue, links become broken, and third-party embedded features can become outdated – just some of many maintenance tasks that accrue over time. ![A screenshot of the Hearing Modernity site before it was maintained by DARTH staff. The domain is expired, and there are styling errors, a broken image, and a disconnected audio plug-in.](/sites/g/files/omnuum11066/files/styles/hwp_1_1__720x720_scale/public/2026-01/Screenshot%202026-01-27%20at%202.12.42%E2%80%AFPM.png?itok=-MgJLAXp) The original Hearing Modernity site demonstrates some of the ongoing maintenance activities: updating domains, resolving security risks, fixing styling issues and broken images, and ensuring third-party plugins remain connected.***Captured with Archive-It (post-maintenance)*** The Hearing Modernity site was a suitable test project for web archiving because it is primarily a custom site that coalesced the outputs from a sound study seminar, including audiovisual materials, links to events, and images. However, you can see below that Archive-It can struggle to capture embedded third-party material, like the streaming audio hosted on SoundCloud, and the featured video. It can also demand more heavy-handed troubleshooting to capture everything in high fidelity. For example, the header is missing in this capture. ![A screenshot of the the Hearing Modernity site captured by Archive-It. Many elements are intact, but the third-party streaming plugins are broken, reducing usability of the project.](/sites/g/files/omnuum11066/files/styles/hwp_1_1__720x720_scale/public/2026-01/Screenshot%202026-01-27%20at%202.19.09%E2%80%AFPM.png?itok=ImeHSyuX) This February 7, 2025 Archive-It capture of the Hearing Modernity project has a missing header and inoperable audiovisual streaming elements. [Link to the Archive-It capture](https://web.archive.org/web/20250207195844/https://hearingmodernity.org/)*.*Archive-It is an excellent choice when you want to “set it and forget it” and maintain a record of changes. For example, if a project fluctuates over time, it is easy to schedule a steady frequency of captures. For archiving and preservation purposes, Archive-It is also the primary tool used across Harvard Library and generates the widely adopted [WARC format](https://www.loc.gov/preservation/digital/formats/fdd/fdd000236.shtml), so it is better integrated into our access and preservation landscape. ***Captured with Webrecorder (post-maintenance)*** Webrecorder is a more hands-on web archiving tool that allows users to copy sites in higher fidelity -- but it also requires more effort up-front, while you essentially scroll through every element that you want to "record.” Below, the same Hearing Modernity site captured by Webrecorder has retained the playability of embedded audiovisual elements. ![A screenshot of the Hearing Modernity site as captured by the Webrecorder tool, showing the styling and streaming plugins are all operable.](/sites/g/files/omnuum11066/files/styles/hwp_1_1__720x720_scale/public/2026-01/Screenshot%202026-01-27%20at%202.22.14%E2%80%AFPM.png?itok=I8NxZ-0n) This Webrecorder capture of the same Hearing Modernity site retains formatting and audiovisual streaming capabilities. [Link to the Webrecorder capture in the DARTH Replay collection.](https://replay.digitalhumanities.fas.harvard.edu/archive/?source=s3%3A%2F%2Fdarth-sunsetting%2Fcrawls%2Fcollections%2Fhearingmodernityfull.wacz#url=https%3A%2F%2Fhearingmodernity.org%2F&ts=20250207212352)Depending on the long-term needs of a project, the higher fidelity record may be worth the additional effort. Other considerations are that Webrecorder can result in larger-sized objects, which can take up valuable storage space. Also, it uses the [WACZ format](https://www.loc.gov/preservation/digital/formats/fdd/fdd000586.shtml), which is a newer and increasingly acceptable preservation format (which *contains* WARC files), but not as widely integrated into the Library’s access pathways. However, For DARTH’s maintenance purposes of static sites, it worked very well. **Now the Hearing Modernity site, as well as other suitable sites, have been sunset and redirected to Webrecorder captures, hosted from the** [**DARTH Replay page**](https://replay.digitalhumanities.fas.harvard.edu/)**.** ### Ephemeral environments for static maintenance This component was an experiment to determine how DARTH could preserve dynamic application user interfaces (UIs), that are otherwise difficult to archive, for data exploration by non-technical users. Ephemeral environments are isolated, temporary deployments of an application that reduce unnecessary infrastructure burden, cloud costs, and security vulnerabilities – and they are often used for testing and validation. The initial plan was to explore commercial products, but they were more targeted at active development and maintenance, as opposed to preserving interactivity for end-users over a long period of time. Cole and Arthur pivoted to exploring “scale to zero” offerings by major cloud providers to achieve the outcome of reducing running and maintenance costs while preserving usability. The [“Visualizing Russian” site](https://visualizingrussian.fas.harvard.edu/) was a suitable test for this experiment because it was a collection of custom, web-based tools with which users can dynamically engage. ![A screenshot of the Visualizing Russian site, which is a Harvard faculty custom collection of web-based tools for studying.](/sites/g/files/omnuum11066/files/styles/hwp_1_1__720x720_scale/public/2026-01/Screenshot%202026-01-27%20at%202.24.59%E2%80%AFPM.png?itok=M0GRnaYk) A screenshot from the Visualizing Russian site. When tested in Google Cloud Run, it maintained its usability and fidelity and took around one minute to fully spin up.In the end, the work to adapt and migrate existing applications to a new cloud provider was cumbersome, and too few applications fit the ideal project profile for this solution. The HUIT partners in our project did recommend to HUIT leadership that alternative strategies, such as Amazon’s similar services for containerized applications, could be explored. ## What happened next? After the pilot wrapped, the team reported their recommendations to HUIT’s ETIP senior leadership team in July 2025. The DARTH website now hosts [the sustainability guidance and dossier template](https://digitalhumanities.fas.harvard.edu/resources/designing-sustainable-projects/), and the [DARTH Replay collection](https://replay.digitalhumanities.fas.harvard.edu/) has allowed select projects to be “sunset” while maintaining their usability. In August 2025, Tricia presented at Harvard Library’s Digital Scholarship Discussion Group, and there will be another opportunity to hear about the pilot in person at the upcoming [“Love Data Rodeo: A Roundup of Useful Ways to Wrangle Your Data”](https://libcal.library.harvard.edu/event/16237169) event in the Lamont Library Forum Room on Thursday, February 12th! See also:- [ Digital Preservation ](/hlps-tags/digital-preservation) Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)