[Paper Review] A Taxonomy and Survey on eScience as a Service in the Cloud

Amelie Chi Zhou, Bingsheng He+1 more|arXiv (Cornell University)|Jul 28, 2014

Scientific Computing and Data Management53 references3 citations

TL;DR

This paper proposes a comprehensive taxonomy and survey of eScience as a Service (eSaaS) in cloud computing, analyzing its application across life sciences, physics, social/humanities, and climate/earth sciences. It evaluates cloud-based eScience against grid-based models, identifies key challenges like cost management, security, and vendor lock-in, and highlights opportunities through national cloud initiatives and improved tooling, calling for cross-disciplinary collaboration to advance the field.

ABSTRACT

Cloud computing has recently evolved as a popular computing infrastructure for many applications. Scientific computing, which was mainly hosted in private clusters and grids, has started to migrate development and deployment to the public cloud environment. eScience as a service becomes an emerging and promising direction for science computing. We review recent efforts in developing and deploying scientific computing applications in the cloud. In particular, we introduce a taxonomy specifically designed for scientific computing in the cloud, and further review the taxonomy with four major kinds of science applications, including life sciences, physics sciences, social and humanities sciences, and climate and earth sciences. Our major finding is that, despite existing efforts in developing cloud-based eScience, eScience still has a long way to go to fully unlock the power of cloud computing paradigm. Therefore, we present the challenges and opportunities in the future development of cloud-based eScience services, and call for collaborations and innovations from both the scientific and computer system communities to address those challenges.

Motivation & Objective

• To develop a structured taxonomy for eScience as a Service in the cloud to systematize understanding of scientific computing in cloud environments.
• To analyze the current state of eScience applications in the cloud across four major scientific domains: life sciences, physics, social/humanities, and climate/earth sciences.
• To compare cloud-based eScience with traditional grid-based computing, identifying advantages and limitations in scalability, cost, and usability.
• To identify critical challenges such as cost optimization, data security, and vendor lock-in in cloud-based scientific workloads.
• To highlight emerging opportunities through national cloud initiatives and improved cloud-native tools, and to advocate for collaboration between scientific and systems communities.

Proposed method

• The authors design a custom taxonomy for eScience in the cloud, covering infrastructure, ownership, application types, processing tools, storage, security, service models, and collaboration mechanisms.
• They conduct a systematic review of existing eScience projects deployed on public and private cloud platforms, focusing on real-world implementations.
• The study compares cloud-based eScience with grid-based computing in terms of performance, scalability, cost, and ease of access.
• It evaluates cloud-native programming models such as MapReduce and DryadLINQ, and workflow systems like DAGMan, for scientific application porting.
• National cloud initiatives—including the US GSA Cloud Storefront, UK G-Cloud, and Japan’s Kasumigaseki Cloud—are analyzed as case studies of institutional cloud adoption.
• The paper synthesizes findings to identify gaps and propose directions for future research and collaboration.

Experimental results

Research questions

• RQ1 How can eScience applications be systematically categorized and understood within the cloud computing paradigm?
• RQ2 What are the key differences and trade-offs between deploying eScience workloads on cloud platforms versus traditional grid infrastructures?
• RQ3 What are the major technical and operational challenges hindering the full adoption of eScience as a Service in the cloud?
• RQ4 What opportunities exist for national and governmental cloud initiatives to accelerate eScience adoption and reduce barriers for researchers?
• RQ5 How can collaboration between the scientific and computer systems communities overcome current limitations in security, cost, and portability?

Key findings

• Despite growing interest, eScience as a Service in the cloud remains in an early stage of development, with eScience tools and systems not yet mature compared to grid-based counterparts.
• Cloud platforms offer significant advantages in scalability, accessibility, and reduced infrastructure management burden, enabling small research groups to handle large-scale data.
• The pay-as-you-go model introduces cost optimization challenges, requiring careful planning to minimize expenses in long-running scientific workloads.
• Security concerns persist due to shared multi-tenancy, necessitating custom security mechanisms for sensitive data, and data lock-in remains a risk due to lack of standardization across cloud providers.
• National cloud initiatives such as G-Cloud, Apps.gov, and Kasumigaseki demonstrate substantial cost savings—projected at £120M by 2015—and promote resource sharing and green data center practices.
• The integration of cloud-native tools like Aneka and SciDB, along with service models such as SaaS and PaaS, significantly reduces development cycles and improves portability of scientific applications.