Research Interests
Mentors
Ph. D in Computer Science
- Conducted research in vehicular edge cloud networks, efficient power management in edge cloud environments, and storage management in multi-tier edge cloud systems. Developed expertise in containers, Docker, and Kubernetes.
- Collaborated with B.Tech and M.Tech students on various research projects.
- Organized various academic talks and conferences, including GCon 2023 and NERC 2023.
M.Tech. in Computer Science and Engineering
- Explored advanced topics in AI, machine learning, and computer vision.
- Achieved rank card holder status for this course.
- Dissertation:
- Focused on brain image segmentation.
- Designed the S-Net model, which optimizes computational complexity by reducing the number of neurons without compromising performance.
- Further refined S-Net to SA-Net by integrating an attention mechanism.
M.Sc. in Computer and Information Science
- Advanced understanding of computer science topics.
- Achieved rank card holder status for this course.
- Dissertation:
- Proposed a novel track placement approach for two- and three-layer channels.
- Enhanced existing techniques to minimize bottleneck crosstalk.
- Successfully simulated the proposed approach, surpassing state-of-the-art performance on benchmark datasets.
B.Sc. in Computer Science
- Acquired foundational knowledge in computer science.
- Organized and led various college and departmental events, including SIGMA 2016, XAVOTSAV 2017, and EXABYTE 2017.
Under Review
Power Aware Container Placement in Cloud Computing with Affinity and Cubic Power Model
Submitted in Future Generation Computer Systems
- Developed an efficient offline container placement strategy on physical machines, accounting for resource constraints, power usage, and machine-container affinity. This work is, to the best of our knowledge, the first to consider these factors comprehensively.
- Introduced a method to derive the affinity between a machine and a container by considering the container’s resource demands and the resources available in the machine.
- Formulated the problem and demonstrated that our approach outperforms state-of-the-art methods on both synthetic and real-life datasets.
Efficient Profit Maximization in Reliability-Constrained Static Vehicular Cloud Systems
Submitted in IEEE Transactions on Services Computing
- Developed an efficient strategy for scheduling tasks to onboard units in parked vehicles at shopping mall parking lots, ensuring real-time processing.
- Achieved 99% task completion reliability within deadlines, significantly improving on the previous best of 50% reliability.
- Introduced unique approaches, including task splitting, task check pointing, and vehicle classification based on historical data.
- Simulated the proposed approach, achieving superior results compared to state-of-the-art methods on both synthetic and real-life datasets.
QoS-Aware Mixed-Criticality Task Scheduling in Vehicular Edge Cloud Systems
Submitted in IEEE Systems Journal
- Proposed an efficient task scheduling approach for vehicular edge cloud systems, addressing task dropping, energy usage, and communication costs. To the best of our knowledge, this is the first work to consider these factors.
- Incorporated mixed-criticality real-time tasks on vehicular edge cloud system, a consideration not previously addressed in the literature.
- Formulated the problem and demonstrated that our approach surpasses state-of-the-art methods on both synthetic and real-life datasets.
Conference
Profit Maximization Using Collaborative Storage Management in Multi-Tier Edge-Cloud Systems
Published in 30th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC), in cooperation with ACM, December 18-21, 2023, Goa, India
- Proposed an efficient data replacement strategy in collaborative multi-tier edge-cloud systems, drawing inspiration from Zipf’s Law.
- Divided the edge storage space into two segments: public and private. The private segment's data replacement is managed by the owner edge server, while the public segment's data replacement is collaboratively managed by all edge servers. This dual replacement policy yielded better results than traditional approaches.
- Formulated the problem and demonstrated that our approach outperforms state-of-the-art methods by a significant margin on both synthetic and real-life datasets.