Eunsam Kim

Disk schedulers for solid state drivers

In embedded systems and laptops, flash memory storage such as SSDs (Solid State Drive) have been gaining popularity due to its low energy consumption and durability. As SSDs are flash memory based devices, their performance behavior differs from those of magnetic disks. However, little attention has been paid on how to exploit SSDs from the disk scheduling algorithm view point. In this paper, we first describe behaviors of SSDs that inspires us to design a new disk scheduler for the Linux operating system. Specifically, read service time is almost constant in an SSD while write service time is not. Moreover, appropriate grouping of write requests eliminates any ordering-related restrictions and also maximizes write performance. From these observations, we propose two disk schedulers: IRBW-FIFO and IRBW-FIFO-RP. Both schedulers arrange write requests into bundles of an appropriate size while read requests are independently scheduled. Then, the IRBW-FIFO scheduler provides complete FIFO ordering to each bundle of write requests and each individual read requests while the IRBW-FIFO-RP scheduler gives higher priority to read requests than the bundles of write requests. We implement these schedulers in Linux 2.6.23, and results of executing our set of benchmark programs shows that performance improvements of up to 17% compared to existing Linux disk schedulers are achieved.

Janus-FTL: finding the optimal point on the spectrum between page and block mapping schemes

NAND flash memory based storage such as SSDs is gaining popularity in commodity computer systems. Some low-end SSDs use the block mapping FTL (Flash Translation Layer) that is good for sequential write patterns but poor for random ones. On the other hand, high-end SSDs tend to use the page mapping FTL that is effective for random write patterns, but whose performance degrades after successive random writes. Designing an FTL that adapts to various workload patterns and provides long-term stable performance is a challenging issue. To resolve this issue, we propose a new FTL, which we call Janus-FTL, that provides a spectrum between the block and page mapping schemes. By adapting along the spectrum, Janus-FTL can provide long-term superior write performance for various workload patterns. We also present a cost model of Janus-FTL that shows the existence of the optimal point on the spectrum for a given workload. Our experimental results show the superiority of Janus-FTL, which adapts itself along the spectrum for a given workload, over state-of-the-art hybrid mapping FTLs and the pure page mapping FTL

Design and implementation of an enhanced personal video recorder for HDTV

We have developed a PVR (personal video recorder) that is capable of simultaneous playback and recording of HD quality MPEG-2 streams for digital TV. It provides viewers with a few enhanced features as well as basic features such as pause, instant replay, skip forward and fast forward/rewind in conventional PVRs for analog TV. This paper describes enhanced and new features that are implemented on our PVR.

Efficient peer-to-peer overlay networks for mobile IPTV services

With recent advance in wireless networks and advent of powerful mobile devices such as smart phones, IPTV services become feasible in mobile platforms. Peer-to- Peer streaming techniques are expected be a promising solution to a large-scale mobile IPTV services by providing low cost and high scalability. On the other hand, the major concern for designing mobile IPTV systems is that mobile devices have the limited battery capacity. To address this concern, in this paper, we therefore develop energy-efficient overlay networks in mobile P2P IPTV systems. In our overlay network for data delivery, mobile peers can reduce the number of data packets and control messages significantly by receiving video data from stationary peers only in a push manner. In our social network for video searching, mobile peers can further reduce a large number of control messages by sending the searching request to only one peer. By extensive simulations, we show that our proposed overlay network structure can decrease the energy consumptions of mobile peers in proportion to the reduced number of data packets and control messages.

Impact of NVRAM write cache for file system metadata on I/O performance in embedded systems

File systems make use of part of DRAM as the buffer cache to enhance its performance in traditional systems. In this paper, we consider the use of Non-Volatile RAM (NVRAM) as a write cache for metadata of the file system in embedded systems. NVRAM is a state-of-the-art memory that provides characteristics of both non-volatility and random byte addressability. By making NVRAM a write cache for dirty metadata, we retain the same integrity of a file system that always synchronously writes its metadata to storage, while at the same time improving file system performance to the level of a file system that always writes asynchronously. To show quantitative results, we develop an embedded board with NVRAM and modify the VFAT file system provided in Linux 2.6.21 to accommodate the NVRAM write cache. The experimental results show that substantial reductions in execution time are possible from an application viewpoint. Another consequence of the write cache is its benefits at the FTL layer, leading to improved wear leveling of Flash memory and increased energy savings, which are important measures in embedded systems.

A scalable PVR-based content sharing architecture for community networks

Integrated with consumer devices like PVRs, a community network can serve as a content sharing architecture for high-quality TV programs in a residential area. To build up such a community network, we employ Fiber Channel Arbitrated Loop (FC-AL) as a fast network connecting a large number of PVRs at a low cost. However, an FC-AL single loop has a critical limitation in terms of the number of attached PVRs when connecting thousands of users in densely-populated communities. We therefore develop a novel multiple loop architecture capable of providing a scalable community network in an economical way without using expensive switches. To connect multiple loops in our architecture, we introduce shared disks that relay TV programs between loops. We then present all possible topologies including linear, ring, edge-added, and complete topology that can be configured with shared disks. We also analyze which topology has the best scalability by evaluating the total traffic of each topology. The analysis and experimental results show that the complete topology among all the possible topologies provides the best scalability.

A performance model and file system space allocation scheme for SSDs

Solid State Drives (SSDs) are now becoming a part of main stream computers. Even though disk scheduling algorithms and file systems of today have been optimized to exploit the characteristics of hard drives, relatively little attention has been paid to model and exploit the characteristics of SSDs. In this paper, we consider the use of SSDs from the file system standpoint. To do so, we derive a performance model for the SSDs. Based on this model, we devise a file system space allocation scheme, which we call Greedy-Space, for block or hybrid mapping SSDs. From the Postmark benchmark results, we observe substantial performance improvements when employing the Greedy-Space scheme in ext3 and Reiser file systems running on three SSDs available in the market.

Optimizations of LFS with slack space recycling and lazy indirect block update

Even though the Log-structured File System (LFS) has elegant concept for superior write performance, it suffers from cleaning overhead. Specifically, when file system utilization is high and the system is busy, write performance of LFS degenerates significantly. Also, cascading update of meta-data triggered by modification of file data decreases LFS performance further. To overcome the performance drawbacks of LFS, we propose two schemes, namely Slack Space Recycling (SSR) and Lazy Indirect Block Update (LIBU). The SSR scheme writes modified data to invalid areas of used segments when on-demand cleaning is inevitable to serve incoming write requests. Also, the LIBU scheme accumulates meta-data update in memory beyond multiple segment writes without compromising consistency so as to decrease total amount of writes. From various experimental results, we observe significant performance improvements when employing the SSR and LIBU schemes for a wide utilization range.

Expediting P2P Video Delivery through a Hybrid Push-Pull Protocol

Despite the recent phenomenal success of peer-to-peer video streaming services, their stumbling performance for high-quality videos remains a major obstacle to wider acceptance. This i ...

Towards greener data centers with storage class memory

Studies have shown that much of todays data centers are over-provisioned and underutilized. Over-provisioning cannot be avoided as these centers must anticipate peak load with bursty behavior. Under-utilization, to date, has also been unavoidable as systems always had to be ready for that sudden burst of requests that loom just around the corner. Previous research has pointed to turning off systems as one solution, albeit, an infeasible one due to its irresponsiveness. In this paper, we present the feasibility of using new Storage Class Memory (SCM, which encompasses specific developments such as PCM, MRAM, or FeRAM) technology to turn systems on and off with minimum overhead. This feature is used to control systems on the whole (in comparison to the previous fine-grained component-wise control) in finer time scale for high responsiveness with minimized power lost to idleness. Our empirical study is done by executing real trace-like workloads on a prototype data center of embedded systems deploying FeRAM. We quantify the energy savings and performance trade-off by turning idle systems off. We show that our energy savings approach consumes energy in proportion to user requests with configurable quality of service. Finally, based on observations made on this data center, we discuss the requirements for real deployment.