- March 30, 2023
- 9:00 AM
- CASE 349 & Zoom
Adnan Maruf is a Ph.D. candidate at the Knight Foundation School of Computing & Information Sciences (KFSCIS) at Florida International University (FIU). He is working as a Graduate Research Assistant at the Data Management Research Laboratory (DaMRL) with his advisor Prof. Janki Bhimani. His research interests are system design, hybrid memory systems, storage systems, and key-value SSD. His research has been published in eight peer-reviewed journals and conferences. He is the first author of three research articles published in highly selective top systems design conferences, such as the IEEE International Symposium on High-Performance Computer Architecture (HPCA’22), The European Event for Electronic System Design and Test (DATE’22), and IEEE International Conference on Cloud Computing (CLOUD’21). Also, his first-author paper is nominated as Best Paper Candidate at DATE’22. He also holds two U.S. patents on designing optimized memory and storage systems. He received the Dissertation Year Fellowship (DYF) award from the University Graduate School at FIU in 2022. He received his B.Sc. in Computer Science and Engineering from Khulna University of Engineering & Technology (KUET), Bangladesh. Before joining the Ph.D. program at FIU, he worked as Sr. Software Engineer at Samsung R&D Institute, Bangladesh, for three years.
Performance and Reliability are two key aspects of the system requirements. Thus, improvement of performance and reliability is one of the most active research areas in system design and development. My dissertation work consists of five innovative prongs. First, the growth of high memory-demanding applications and emerging technologies, such as persistent memory (PM) that have high performance, low cost, and low power consumption, motivates to design of high-performing yet low-cost systems. We design Multi-Clock, an efficient, low-overhead hybrid memory system that relies on a unique page selection technique for tier placement. Second, we discuss the importance of allocation policies in tiered memory systems. We present how significantly the allocation policies can impact the dynamic tiered memory systems. Third, we develop a memory access tracing (MAT) kernel to collect and analyze workload behaviors and tiered memory systems characteristics. MAT-kernel can produce important data regarding the page allocation, page access, page migration, and page scan. Fourth, due to the rapid adoption of flash-based solid-state drives (SSDs) as the main storage media, popular and contemporary applications such as console gaming require in-depth investigation to optimize the storage and application performance. We take the first attempt to systematically measure, quantify, and characterize the organization of game data into files, back-end storage access patterns, and the performance of gaming workloads. Finally, the widely deployed SSDs and the global climate change warnings drive the exploration of the reliability of such storage devices under extreme environmental conditions. We design and perform methodical exploration to analyze, measure, and observe the impact of various commonly changing environmental conditions, such as temperature and humidity, on the performance of SSDs. Overall, in this dissertation, we leverage the key benefits of emerging technologies to design and improve system performance. We perform in-depth exploration to understand the evolving applications and the reliability of modern storage systems.