YJYuhun Jun

Yuhun Jun

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Yuhun Jun has a total of 6 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    An Adaptive Zone-Grouping Scheme Enabling General-Purpose File Systems on ZNS SSDs

    by Jungyun Choi, Yuhun Jun, Jongseok Kim, Jinkyu Jeong, Euiseong Seo

    Zoned namespace solid state drives (ZNS SSDs) provide superior capacity and life span compared to conventional namespace (CNS) SSDs at the same cost. Especially, a small-zone ZNS SSD can achieve a low write amplification factor (WAF) and performance isolation among I/O streams by providing the host with finer-grained control of the SSD's flash chips than the large-zone counterpart. To maximize these advantages, multiple zones in a small-zone SSD should be grouped and operated simultaneously. A ZNS-aware file system encapsulates the ZNS interface and transforms a ZNS SSD into a general-purpose storage to the host layer. However, the current ZNS-aware file systems were designed for the large-zone ZNS SSDs, and thus lack the zone group management feature. This paper proposes a dynamic zone group management scheme for ZNS-aware file systems that dynamically adjusts the number of zone groups in accordance with the workload characteristics and the size of each zone group to achieve high performance at a low WAF. The proposed scheme was implemented in the F2FS. Our evaluation with FIO, Filebench and db_bench showed that, across all cases, the proposed scheme improved the performance by up to 71.2% while reducing the WAF by 29.8% in comparison to the static zone grouping scheme.


  • Open Access

    Analysis and Mitigation of Patterned Read Collisions in Flash SSDs

    by Yuhun Jun, Jaehyung Park, Jeong-Uk Kang, Euiseong Seo

    A modern flash solid-state drive (SSD) achieves superb throughput by accessing its flash memory dies in parallel. To obtain parallelism in processing writes, the flash translation layer (FTL) of an SSD is allowed to allocate physical pages from idle dies or dies with low loads. However, since the die that holds the page to be read is determined in advance, when multiple read requests head for one die, read collisions occur, and the read operations involved in the collisions must be serialized. These read collisions lead to a significant prolongation of read latency and, thus the degradation of throughput. As the density of flash dies increases, more pages are stored on a die, which is expected to result in more frequent read collisions. Unlike the write collisions that the DRAM buffer can absorb, read collisions directly affect user-experienced latency. Therefore, the severity of the problem is further increased. In this paper, we identify the impact of read collisions on performance with real-world traces. We also propose an approach that distinguishes patterned read collisions from accidental ones and mitigates the performance impact of patterned read collisions. By replicating frequently occurring pages involved in patterned read collisions, the proposed approach improves the average and tail read latency of an SSD. The evaluation through simulation with the 34 MSR Cambridge traces showed that 12 traces out of 34 earned read performance improvement larger than 10% and up to 37%. In addition, the tail read latency of 15 traces was reduced by at least 20% and by up to 53%. Only two traces showed negligible degradation in average and tail read latency by around 1%.