Random Access Concatenated Libraries and dd enable a short-latency high-content website on an inexpensive shared server

Abstract

Content-rich websites typically house their images as individual files or as more costly binary database objects. Both methods place high demands on storage resources with commensurate high monetary cost. Inexpensive shared service accounts come with strict limits on disk and computing resources which prevent their use for hosting content-rich websites. To minimize the load on these limited resources, large numbers of images or other records may be concatenated into a single file and delivered with short latency independent of location within the file using Linux utility dd. This solution and its performance are presented for (a) a website which houses 8000+ genealogical reference books with more than 3.5 million 250-Kbyte page images, (b) a website-based database of 86 million death records, and (c) a cluster computing application which utilizes 800 neuroimaging databases, each with 1.5 million records. These disparate applications demonstrate the efficacy, scalability, and general utility of the solution.