Pipeline-Oriented Web and Grid Computing Frameworks for Large-Scale E-Learning Systems

Abstract

Massive-scale e-learning is beginning to be run with conditions of massive co-located access, heterogeneous workloads and dynamism of resource demands, dramatising underlining scalability and automation constraints of traditional centralised and monolithic web-based learning management systems. In order to overcome these issues, this paper presents a pipeline-based web and grid computing architecture of scalable e-learning systems. The essence is to provide the ability to perform high throughput, automated implementation of e-learning processes by introducing explicit structure of modelling learning processes as distributed stages of the pipeline. The offered architecture breaks down end-to-end e-learning processes - starting with content transmission and recording of interactions to analysis of evaluation results and analytics - into sequential and parallelizable pipeline processes. Service interfaces (which are web-based) process user interactions, and pipeline stages which require significant computation are dynamically mapped back to distributed grid computing resources. A pipeline execution and scheduling model is established that optimises the allocation of tasks, orchestration of data flow and the utilisation of resources using pipeline parallelism. The framework is enforced with the help of modular services and workflow orchestration layer, which facilitates concurrent scheduling and fault separation. The performance assessment is done under artificial large workloads and using different amounts of concurrent learners and a nonhomogeneous compute hardware. The experimental evidence shows that the suggested pipeline-based framework can easily attain significant enhancements in the system throughput, response latency and resource utilisation, compared to non-pipelined centralised architectures, especially at peak load. The findings affirm that pipeline-based automation and grid execution is a scalable and efficient base used to support next-generation distributed e-learning platforms. The paper underscores the topicality of pipeline-based distributed computing as an effective architectural model of a large-scale e-learning system.