Excellent Paper Award at the Korea Software Congress 2020
Files
The paper 'A Container-Based Scalable Fabric-Attached Memory Management Platform', written by graduate student Lee Hee-rak of the Department of Computer Science & Engineering (master's, 2nd semester) under the supervision of Professor Kim Young-jae of the same department, received the Excellent Paper Award in the computer systems field at the Korean Institute of Information Scientists and Engineers conference (KSC 2020) held on 21 December 2020.
▲ From left: Lee Hee-rak (master's, 2nd semester; supervisor: Professor Kim Young-jae), Lee Chang-gyu (combined master's–doctoral, 7th semester; supervisor: Professor Kim Young-jae),
Awais Khan (doctoral; supervisor: Professor Kim Young-jae), Kang Hyun-gu (combined master's–doctoral, 4th semester; supervisor: Professor Kim Young-jae)
▲ Certificate from the Korean Institute of Information Scientists and Engineers conference
[Figure 1] A case of data consistency failure caused by the direct accessibility of Fabric-Attached Memory. In the area marked with a red dotted line, data consistency may not be preserved because of simultaneous access by two nodes.
[Figure 2] An example of the memory management protocol provided by the container-based framework (memory region allocation)
With the arrival of the big data era and the growth in the data systems must process, research into memory-centric computing to overcome capacity limits and process data at speed has drawn attention. Fabric-Attached Memory (FAM) is a high-performance memory technology devised for memory-centric computing systems. In a shared memory environment composed of FAM, when a particular computing node accesses a memory region it can access all memory regions directly without communicating with other computing nodes, unlike in conventional shared memory environments.
This research resolves the data inconsistency problem that can arise from direct access to memory regions in a FAM-based shared memory environment, while proposing an efficient container-based memory management framework for exploiting high-performance FAM. The framework allocates an independent memory region to each application requiring FAM, and provides an efficient memory management protocol that can handle simultaneous requests from multiple applications with low overhead.
The Korean Institute of Information Scientists and Engineers conference (KSC 2020) is an information science and technology conference at which the latest research results across every area of information technology — artificial intelligence, computer systems, information security and high-assurance computing and more — are presented.