A Provisioning Service for Automatic Command Line Applications Deployment in Computing Clouds

Abstract

This paper discusses an approach for deploying command-line applications on cloud computing platforms. The main focus of the paper is on execution of scientific algorithms (such as algorithms from the music information retrieval domain) by enabling the sharing of algorithm implementations, the reproducibility of results, and an easy access to non-experts of distributed environments. We propose a method of automatic deployment of client command line applications in PaaS and IaaS clouds which, unlike to existing approaches, requires neither preliminary platform configuration, nor platform configuration description. We designed a deployment manager using an expert system in order to automatically setup a virtual execution platform, by resolving dependencies and solving configuration issues on the basis of a pre-existing knowledge base.

Approach to Building a Web-based Expert System Interface and Its Application for Software Provisioning in Clouds

“…6 According to Section III-A a context for a Request is an Activity. By convention a top-level activity may represent a user and a context for all top-level requests.…”

“…As we described in our earlier work [6], we extended the MEDEA and NEMA execution model by adding the second phase of the execution process. During the first phase (command execution) some debugging information can be written to stdout/stderr, to the environment logs and so on.…”

“…For each error we provided the knowledge base rule in order to detect and fix configuration errors. Table III lists 3) Change Scenario: Managed and Unmanaged Execution Modes: As we described in [6], the architecture with isolated proxy and executor components allows implementing different execution modes (e.g. system behavior) without modifications of neither the proxy, nor the executor.…”

“…One significant problem is a relatively high barrier to entry for non-experts. In our earlier work we described sources, advantages and problems of deploying research software in clouds and proposed an architecture of a provisioning service for automatic CLI applications deployment in computing clouds [6]. The proposed solution targets problems of build and run error discovery and handling, with special emphasis on errors conditioned by possible misconfiguration of a virtual platform where client software modules have to be executed.…”