Olive: Sustaining Executable Content Over Decades

We describe a system called Olive that freezes and precisely reproduces the environment necessary to execute software long after its creation. It uses virtual machine (VM) technology to encapsulate legacy software, complete with all its software dependencies. This legacy world can be completely closed-source: there is no requirement for availability of source code, nor a requirement for recompilation or relinking. The entire VM is streamed over the Internet from a web server, much as video is streamed today. 1. Software in Science Reproducibility is at the heart of the scientific method. Confidence in a result grows as researchers all over the world are able to reproduce it independently. Today, an increasing fraction of the world’s intellectual output is in the form of executable content — i.e., software. This is true in virtually all areas of scholarship, from physics, chemistry, biology, and engineering to economics, political science and the humanities. Examples of such executable content include data analysis tools to slice and dice raw data, zoomable visualization tools that enable results to be viewed at many levels of abstraction, and simulation models written in a variety of programming languages and using a wide range of supporting libraries and reference data sets. Such software is central, not peripheral, to the discovery of new results today. Raw scientific data is often of limited value unless it is accompanied by the uniquely customized software that was created to decode, interpret, analyze and display that data. The role of software in the scientific method is illustrated by a recent controversy [6]. In early 2010, Reinhart and Rogoff published an analysis of economic data spanning many countries [8, 9]. Herndon et al [4] refuted their findings in 2013 by discovering an error in their calculations. The significance of the error was described as follows [7]: “The Reinhart-Rogoff research is best known for its result that, across a broad range of countries and historical periods, economic growth declines dramatically when a country’s level of public debt exceeds 90 per cent of gross domestic product. · · · When we performed accurate recalculations using their dataset, we found that, when countries’ debt-to-GDP ratio exceeds 90 per cent, average growth is 2.2 per cent, not -0.1 per cent.” The controversy continues, but regardless of how it is eventually resolved, there is no denying the central role of software (in this case, a Microsoft Excel spreadsheet) in the original analysis, its refutation and its eventual resolution. 2. The Ravages of Time In the Reinhart-Rogoff example, there was no difficulty in obtaining the software necessary to perform the recalculations. Only three years had elapsed since the original publication of results, and the same version of Microsoft Excel continued to be in widespread use. Imagine, however, that the recalculations were attempted by a researcher 30 years later. Would Microsoft Excel still be in use? If so, would the version then in use accept the data format used by the original researchers? Would the calculations performed by that version be identical in every respect (including, for example, handling of rounding errors) to the version used by the original researchers? What if Microsoft goes out of business ten years after the original publication of results, and the Windows environment (which is needed to run Excel) ceases to be in use? As these questions suggest, our growing dependence on software in scientific research introduces new challenges to the premise of reproducibility that is the bedrock of science. Unless these challenges are addressed, our ability to re-validate published results will evaporate over time. In this paper, we describe a system called Olive that seeks to freeze and precisely reproduce the environment necessary to execute software long after its creation (possibly many decades later). It uses virtual machine (VM) technology to encapsulate legacy software, complete with all its software dependencies. This includes the operating system, dynamically linked libraries, tool chains, configuration files, data files and other supporting items. This legacy world can be completely closed-source: there is no requirement for availability of source code, nor a requirement for recompilation or relinking. The entire VM is streamed over the Internet from a web server, much as video is streamed today. Oneclick execution of pre-packaged legacy software from a web site thus becomes possible. The rest of this paper examines the challenges addressed by Olive, and then describes its design, implementation and current status.