There are a lot of choices to make when implementing a IIIF server. Does one use TIFF or JP2 images as the source image format? If JP2s, should Kakadu or OpenJPEG be used? What about where the server should run? What are the advantages of running a server locally vs. in the cloud on EC2, or as a Docker container on AWS' infrastructure? There are also lots of caching options that one might want to explore, not to mention the possibility of using newer services like AWS Lambda to scale the work. How does an institution make these decisions?
At UCLA we decided to experiment with the many options available and to come up with a series of metrics that we could use to inform our decisions. For instance, what sort of impact does the caching layer have on the time that it takes to serve a newly requested (not cached) image on our IIIF server? Does the upload time for source images outweigh the value of running things at scale on Lambda? We propose to talk about our experiments, the metrics we've gathered, and the conclusions we've come to as a result of our process.
