Workflow as a Service Broker in Cloud Environment: A Systematic Mapping Study

Saeid Abrishami, Faridreza Momtaz Zandi, Alireza Nourbakhsh · 2025

Read Workflow as a Service Broker in Cloud Environment: A Systematic Mapping Study by Saeid Abrishami, Faridreza Momtaz Zandi, Alireza Nourbakhsh (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-12672.txt.

Introduction

Saeid Abrishami, Faridreza Momtaz Zandi, and Alireza Nourbakhsh map the research on cloud workflow brokers in a twenty twenty-five study. Scientists run complex workflows in the cloud, but choosing resources, prices, and schedules is a heavy burden. Workflow as a Service brokers take that burden over.

The team ran a systematic mapping study with a three-tier search: database queries plus backward and forward citation chasing. Eighty-seven strong papers from forty-nine venues passed their screen. They sort the field by broker architecture and answer five research questions. The map names the main trends, the open problems, and where broker design should go next.

Catalog record

CREATORSaeid Abrishami, Faridreza Momtaz Zandi, Alireza Nourbakhsh
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE120 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-12672

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Workflow as a Service Broker in Cloud Environment: A Systematic Mapping Study Saeid Abrishami, Faridreza Momtaz Zandi, Alireza Nourbakhsh Abstract Cloud computing has emerged as a promising platform for running scientific workflows across various domains. Scientists can take advantage of different cloud service models, such as serverful or serverless, to execute workflows based on their specific requirements, along with diverse pricing models like on-demand, reserved, or spot instances to reduce execution costs. However, the challenge of selecting appropriate resources and pricing models, coupled with the orchestration and scheduling of workflow tasks, creates significant complexity for users. To mitigate this burden, Workflow as a Service (WaaS) brokers have been introduced to facilitate workflow execution. In recent years, numerous studies have been published, either directly or indirectly related to this research area, highlighting the need for a comprehensive and systematic review of WaaS brokers to identify key trends and challenges in this field. In this paper, we conduct a Systematic Mapping Study (SMS) on WaaS brokers within cloud environments. The SMS employs a thorough 3-tier strategy (database search, backward snowballing, and forward snowballing) to answer five research questions. A total of 87 high-quality articles, published in 49 prestigious venues, are analyzed to derive a taxonomy based on the architecture of WaaS brokers. The articles are classified and surveyed according to this taxonomy, and future research directions for the design and implementation of WaaS brokers are explored. This study provides valuable insights for researchers and developers, helping them identify major trends and issues in the field of WaaS brokers. Journal: Computer Science Review † † highlights: Conducts the first systematic mapping study on Workflow-as-a-Service brokers. Develops a structured taxonomy for WaaS broker components and architectures. Analyzes 87 primary studies to reveal prevailing trends and open research issues. Outlines future research toward AI-driven, serverless, and hybrid quantum/classical brokers. 1 Introduction Cloud computing has become one of the most transformative technologies of the modern era, reshaping industries and revolutionizing how computing resources are accessed and utilized. It enables customers to rent resources on demand and pay on a pay-as-you-go basis, eliminating the need for large upfront investments in infrastructure [ 85 ] . Both industry and academia have widely adopted this computational paradigm to meet their growing and diverse resource requirements [ 169 ] . In particular, emerging paradigms in artificial intelligence (AI)—such as deep learning and large language models—benefit from the cloud’s ability to provide scalable access to resource-intensive infrastructure like GPUs. In response to evolving user demands, leading cloud service providers (CSPs) continue to innovate by offering new services aimed at improving usability, flexibility, and performance. As a result, beyond traditional offerings like Infrastructure as a Service (IaaS), which provides virtual machines (VMs), a range of higher-level services has emerged. These include Container as a Service (CaaS), Function as a Service (FaaS)—also known as serverless computing [ 146 ] —and Quantum as a Service (QaaS), also referred to as Quantum Computing as a Service (QCaaS) [ 95 ] , each designed to abstract away infrastructure management and streamline application deployment. Scientific workflows are widely used to automate and execute complex scientific applications across diverse domains [ 86 , 2 ] . A scientific workflow defines a process aimed at achieving a scientific objective, typically represented as a set of tasks and their dependencies. To support the composition, planning, orchestration, and automated execution of such workflows, researchers commonly rely on Workflow Management Systems (WMSs) [ 158 ] —such as Pegasus [ 46 ] , Apache Airavata [ 56 ] , and Apache Airflow [ 57 ] . These applications usually require resource-rich computing environments, such as distributed infrastructures. Cloud computing offers an attractive solution, enabling scientists to exploit services such as IaaS, CaaS, and FaaS, together with pricing models like on-demand , spot , or reserved , to execute workflows efficiently and cost-effectively. A central challenge, however, lies in selecting appropriate resources and pricing models and, more critically, orchestrating and scheduling workflow tasks across them. This is where cloud brokers become essential. Brokers mediate between CSPs and customers, identifying suitable services while incorporating user preferences [ 90 ] . They also mitigate the vendor lock-in problem , i.e., dependence on a single CSP’s services, which may otherwise lead to monopolistic scenarios. Workflow as a Service (WaaS) represents a class of cloud brokers that accept users’ workflows together with their specified Quality of Service (QoS) requirements, such as deadline and budget , and execute them by provisioning and allocating suitable resources from one or more CSPs. While the broker seeks to maximize its own benefits, it must also provide sufficient incentives for users to outsource their workflows, most notably by reducing execution costs. To accomplish this, the broker relies on two primary components: the resource provisioner , which is responsible for selecting resources across different service and pricing models, and the task scheduler , which maps workflow tasks onto the provisioned resources. By executing multiple workflows concurrently, the scheduler can improve overall resource utilization compared to scheduling workflows independently. This makes scheduling in WaaS brokers a distinctive challenge, commonly referred to as the Multiple Workflow (or Multi-Workflow) Scheduling Problem (MWSP) . A structured review of WaaS brokers is crucial for uncovering trends, challenges, and gaps in this domain. To achieve such coverage, systematic methodologies are typically adopted, most notably the Systematic Literature Review (SLR) [ 92 ] and the Systematic Mapping Study (SMS) [ 127 ] . While both approaches share a rigorous procedure for locating, filtering, and selecting studies, their objectives are fundamentally different. An SLR is a focused, evidence-oriented method designed to synthesize and compare existing research in depth, often including a critical evaluation of methods, metrics, and performance results under similar conditions. By contrast, an SMS does not aim to compare or rank methods. Instead, its role is broader and more exploratory: it maps the landscape of existing work, categorizes studies by dimensions such as method, contribution type, or evaluation strategy, and analyzes descriptive statistics such as publication years, venues, and frequency of techniques. The emphasis is on identifying research trends and highlighting underexplored areas, thereby revealing gaps in the literature. In this sense, SMS research questions are intentionally general, seeking to provide a big picture of the field, whereas SLRs drill down into fine-grained, evidence-based comparisons. An SMS often lays the groundwork for one or more SLRs by highlighting sub-areas that merit detailed comparative study. To the best of our knowledge, no prior SMS has been conducted specifically on WaaS brokers. While a few studies address the broader domain of cloud brokerage [ 106 , 90 ] , they do not specifically examine WaaS brokers. However, given that the scheduler is a key component of a WaaS broker, surveys on workflow scheduling in cloud environments are considered related works. Several studies have surveyed cloud workflow scheduling in general [ 169 , 3 , 94 , 116 ] , but while these articles touch on the MWSP, they do not comprehensively address it. Recently, Saeeidzade and Ashtiani [ 140 ] published a survey on workflow scheduling in cloud environments, which includes a subsection on scheduling in WaaS brokers. However, this survey does not comprehensively explore the domain. The most relevant work to

Read the complete text