Thompson, Christopher, Ukkusuri, Satish, Gehlot, Hemant, Lei, Zengxiang, Verma, Rajat, Xue, Jiawei, Qian, Xinwu, Xianyuan, Zhan, Rice, Shawn, and Song, Carol
While there is no shortage of research teams developing complex modeling code across the spectrum of scientific domains, most are not prepared to transition their simulations from private labs into publicly accessible gateway platforms. Research software engineers (RSEs) can bridge this vital knowledge gap, and in doing so help researchers both reach a greater audience for their work and build a greater value returned on their research funding. This presentation details the case of Purdue University RSEs helping build a gateway to the A-RESCUE simulation[1], a realistic vehicle traffic model for studying how efficiently an entire population will evacuate through a city-scale road network. Each project has unique challenges that will place constraints on what solutions are viable. RSEs for this project balanced what is technically the best direction for the architecture against the future roles of the researchers. They will be maintaining the gateway independently after the collaboration period, operating within the computing environment they have already created, and staying within the bounds of their grant deliverables. This project began by evaluating the existing simulation interface. The research team had developed a robust model which operated solely through a framework vendor provided GUI. This ran in a standard workstation but required each end user to setup a development environment and run the model through a specialized tool. For potential research collaborators this may be viable, but it would set too high a bar for any potential lay-user such as emergency personnel wanting to plan city evacuations. RSEs were able to raise these concerns and propose the creation of a new server interface built around an alternate batch-mode API found within the framework’s documentation. This allowed the simulation to be made available to new classes of remote users through a simple network socket with almost no modification of the underlying simulation code and no local software installation necessary. A-RESCUE is written in the Java language using the Repast Simphony agent-model framework[2]. This factored into the solution platform choice as the researchers would need to self-maintain the gateway after development concluded. NodeJS [3] was selected as the best platform for the gateway as it has excellent community support to help the team with debugging future issues, is written in JavaScript at which they were already proficient, and meshed well with their model Java code. With the server solution determined, the RSEs aided the team with refining a prototype web client and composing all the systems together within servers they had already purchased for the grant. Working with campus computing staff, they gained access to additional resources for server management and professional UI development for their web client prototype. Research teams utilizing campus RSEs to build scientific gateways adds value for all involved. Rather than merely optimizing model code, RSEs can create interfaces between applications, introduce new technologies, and streamline workflows for easier maintainability. These collaborations generate new audiences for both the research team’s work and the institution’s computing center. Promoting these collaborations should be a priority to all organizations.