PURR was founded in 2011 as a partnership between Purdue University Libraries, Information Technology at Purdue (ITaP), and the Office of the Executive Vice President for Research as campus-wide support for researchers throughout the data management lifecycle. It was built on the HUBzero(r) platform, which was developed at Purdue. PURR provides the tools and expertise to help researchers plan for data management, share data with collaborators, publish completed datasets in compliance with federal funding guidelines, safely archive data, and track data publication impact. Over the past eight years, PURR has published 975 datasets, and served over 3,600 researchers with 481 grant awards. In that time, PURR's services have grown along with the HUBzero(r) platform to meet the changing needs of the Purdue community as researchers across all fields produce more data. Supporting larger datasets requires a multi-faceted approach far beyond simply acquiring additional storage space. Our recent development has followed a 5-pronged plan: 1) increased storage quotas, 2) new publication series functionality, 3) an online database viewer, 4) publication file preview, and 5) seamless ftp transfers for large publications. Combined, these improvements ensure our increasingly large data publications are not only stored safely, but also are accessible over the long term. Although these developments were designed for PURR, many of them are now incorporated into the open source HUBzero(r) core code available to other institutions. They may also serve as a model for other small science gateways that are working with increasingly large datasets. The newly published Rough Cilicia Survey Pottery Study dataset series illustrates both the motivation for and the results of PURR's recent developments. The culmination of four years of close collaboration between PURR's data curator and a faculty member from Purdue's classics department, the Rough Cilicia collection is composed of 25 datasets. This series makes available images of hundreds of pottery sherds from the ancient Cilicia region of modern-day Turkey, and their associated descriptive information in a series of interactive data tables that allow the user to view, search, and filter data on the PURR website. Users can also download the data files for closer study and reuse. At about 15 GB, the Rough Cilicia series is not exactly "big data," but it is large enough to stretch the limits of a web-based repository like PURR, and we are increasingly seeing datasets of this size or more. Moderate improvements like the five mentioned here allow us to publish larger datasets while maintaining the ease and convenience of serving users through a web browser.