Shu Yin's research while affiliated with ShanghaiTech University and other places

... While the file system is running at the mount time, the FUSE daemon cannot be called by VFS without the use of FUSE. In some cases, FUSE may take another action, such as returning the requested data into the buffer, or it may perform some pre-processing by requesting data from the underlying file systems [18]. Eventually, when the mounting connection is no longer needed, the mount point directory contents automatically disappear as soon as the file system is unmounted. ...

... For example, attacks on wireless mobile ad hoc networks are investigated in [15], attacks on wireless sensor networks [16], and cloud computing environments are discussed in [17]. Furthermore, due to a lack of end-to-end novel security solutions, several attacks have been against evolving technologies, for instance, those depending on artificial intelligence (AI) [18], augmented reality (AR) [19], Software-Defined Networking (SDN) paradigm [20] and Blockchain addressed in [21]. Attacks against new business models and e-health applications presented in [22]. ...

... centimeter to decameter and beyond), coupled with the computational overhead of operating VR user interfaces with large textures or vertex counts (e.g. Lyu et al., 2019) offers further challenges in terms of latency. To capture heterogeneities at multiple length scales such that VR user experiences approach those of the field, it is clear that visualizations using locally varying texture resolution and/or vertex densities (e.g. ...

... Both for live and non-live migration schemes, there is a need for efficient resource allocation to carry out the migration process [34]- [37]. Migration can be performed either using a wired or wireless network. ...

... In HDDs, data manipulation may significantly affect the performance due to disk rotation to access different tracks. Although SSDs have better performance and the cost per gigabyte has decreased, solid-state drives are still more expensive than HDDs (about 33.33% [30]). Solid-state drives are devices that have a similar block I/O interface to hard disks [31]. ...

... Results indicate energy consumption is responsible for 5 to 28% of overall cost for databases placed in HDDs. [20] ✳ ✳ ✳ Tan et al. [14] ✳ ✳ ✳ Wan et al. [21] ✳ ✳ ✳ Mingzhou et al. [22] ✳ ✳ Salkhordeh et al. [23] ✳ ✳ Desnoyers et al. [24] ✳ ✳ Li et al. [25] ✳ ✳ Boukhelef et al. [11] ✳ ✳ ✳ ✳ Jiao et al. [16] ✳ ✳ ✳ Kim et al. [26] ✳ ✳ ✳ Jingyu et al. [27] ✳ ✳ ✳ Boukhelef et al. [15] ✳ ✳ ✳ Wu et al. [28] ✳ ✳ ✳ ✳ Lin et al. [29] ✳ ✳ ✳ ✳ ...

... When the data is prefetched using uniform distribution and Zipf distribution (SP_Uniform, Sp_Zipf, and Sc) [29,30], the average cache hits are calculated as 56.2%, 55% and 46.6%, respectively. Our present technique outperforms the others by providing an average cache hit ratio of 78.5%. Figure 10b compares the execution time taken by our proposed PAP framework with the Reliable Energy-Efficient Storage System (RESS) and Modified Parallel Log-Structured File System (PLFS) proposed in [31]. The results demonstrate that PAP performs much better for 100 and 200 queries by taking 23 and 30 s, respectively. ...

... Yin et al.propose architecture for active storage systems and a hybrid disk configuration for the storage system in active storage. Termed as the HcDD the disk architecture consist of dual buffer; [8] a HDD write buffer for deduplication of write requests and SSD buffer for providing parallelism to write requests. Read requests are directly processed by the SSD whereas write request to an active storage system is sent to the HDD buffer for deduplication before processing. ...

... 90% of all data in the world was generated over the last 2 years [4], and global data are predicted to reach 163 zettabytes by 2025 [5]. Facebook and Google respectively process 20 [6] and 25 petabytes of data per day [7]. ...

... The concept of pipelined encoding process to improve archival system is explored in [12]. A novel system for archival storage [13], novel erasure codes concept [14], a data placement scheme in distributed environment for dataintensive applications [15], decentralized erasure coding for efficient data archival system [16], erasure coding with parity logging in cluster storage environment [17], the utility of Software Defined Network (SDN) to decouple data and control planes in distributed environments, and graph based optimizations in Hadoop framework for improving functions like Swappiness and Hugepage [18] are other important contributions found in the literature. ...