Research
I conduct research in multiple domains in computer systems including cloud and edge computing, parallel and high performance computing, performance engineering, and system security, among others. Current I am supervising multiple Final Year Undergraduate Research Projects, MSc Research Projects as well as MBA Research Projects. Furthermore, I am a founder and a key member of the Complex Systems Research Lab - CysResLab at the Department of Computer Science and Engineering, University of Moratuwa.
Final Year Undergraduate Research Projects
2022-2023
Trusted FaaS Computing
Avishka Shamendra, Gayangi Seneviratne, Binoy Peries
This project intends to propose tools and techniques for ensuring the trustworthyness of FaaS computing applications. The project will involve extending an open source FaaS cloud platfrom such as OpenWhisk or OpenFaaS to facilitate integrity preservation and verification mechanisms for a FaaS cloud user and a remote attestor. Keywords: Trusted Computing, Cloud Security, Virtualization and Containers
Achieving Energy Efficiency Through Containerization: A Brownout Approach
Lakshan Banneheke, Dhaura Pathirana, Thushani Jayasekara
Containers have emerged as a key technology enabler, especially with the adoption of microservices architecture in developing cutting edge software solutions. Container orchestration refers to the automation of managing containerized workloads. Brownout approaches rely on disabling/enabling system and software components dynamically to achieve various system and application objectives. In this research, we investigate container orchestration systems coupled with a brownout approach for achieving energy efficiency. This research will also involve developing workload prediction strategies, machine learning based or otherwise, to speculate application demand in future. Keywords: Virtualization, Containerization, Container Orchestration, Energy Efficient Computing, System Performance, Machine Learning
Cost Optimization on Hybrid Clouds
Dasith Edirisinghe, Kavinda Rajapakse, Pasindu Abeysinghe
This project will look into optimizing the cost of hybrid clouds by optimizing energy consumption in the private clouds and using spot instances in the public cloud, avoiding Service Level Agreement (SLA) violations caused by abrupt terminations of spot instances.
Cost-time performance of serverless hybrid cloud applications
Prathees.P, Pirathees.K, Dilaxsaswaran.N
This research project investigate the dynamics between cost, time, and application performance of serverless applications in a hybrid serverless cloud infrastructure. We develop measurement-driven analytical models to predict the performance of serverless applications, validate the models, and propose recommendations for efficient resource allocation for achieving various application objectives.
Self-healing and self-adaptive management of IoT-based edge computing infrastructure
Sahan Samarakoon, Shashika Bandara, Nishan Jayasanka
Co-supervised by Dr. Chathuranga Hettiarachchi
With current fault-tolerant techniques and architectures not addressing the self-adapting and self-healing concerns for edge computing environments, uers often needs to rely on manual intervention and significant downtime when there is a fault. In this project, we investigate tools and techniques to identify faults in the edge computing environments and correct them automatically with minimal human intervention.
2021-2022
Syncbox Cloud Accelerator
Udara Pathum, Evindu Rajapaksha, Melanga Kasun
Although cloud services provide good features, high performance, low cost, and reliability, the availability, speed, and cost of the Internet connection may be an issue while accessing cloud services. This may reduce the productivity of organizations and work-from-home employees. SyncBox provides uninterrupted cloud services during network outages and improves the quality of cloud services by caching and synchronizing files between the main file server and the device. Furthermore, we implemented a deep learning model using Temporal Fusion Transformers to perform time-shifted prefetching to increase file access speed by shifting a part of the data transfer to the off-peak time where the network traffic is low. According to our benchmarks, using cloud services through SyncBox was 3.5 times faster than accessing cloud services directly from the internet. We evaluated the prefetching and file caching process, and the results revealed that users were able to access more than 58% of the files from the cache without experiencing any network disruptions and with higher throughput.
MSc Research Projects
MBA Research Projects
- Malith Samarathunga, How Cloud Adoption Affects System Administration Professionals in Sri Lanka, 2022