Kern Koh

Efficient replacement of nonuniform objects in Web caches

Cache performance depends heavily on replacement algorithms, which dynamically select a suitable subset of objects for caching in a finite space. Developing such algorithms for wide-area distributed environments is challenging because, unlike traditional paging systems, retrieval costs and object sizes are not necessarily uniform. In a uniform caching environment, a replacement algorithm generally seeks to reduce cache misses, usually by replacing an object with the least likelihood of re-reference. In contrast, reducing total cost incurred due to cache misses is more important in nonuniform caching environments. The authors present the least-unified value algorithm, which performs better than existing algorithms for replacing nonuniform data objects in wide-area distributed environments.

A flash compression layer for SmartMedia card systems

Flash memory based SmartMedia card is becoming increasingly popular as data storage for mobile consumer electronics. Since flash memory is an order of magnitude more expensive than magnetic disks, data compression can be effectively used in managing flash memory based storage systems. However, compressed data management in flash memory is challenging because it only supports page-based I/Os. For example, when the size of compressed data is smaller than the page size, internal fragmentation occurs and this degrades the effectiveness of compression seriously. In this paper, we developed a flash compression layer (FCL) for the SmartMedia card systems. FCL stores several small compressed pages into one physical page by using a write buffer. Based on prototype implementation and simulation studies, we show that the proposed system offers the storage of flash memory more than 140% of its original size and expands the write bandwidth significantly.

A fast start-up technique for flash memory based computing systems

Flash memory based embedded computing systems are becoming increasingly prevalent. These systems typically have to provide an instant start-up time. However, we observe that mounting a file system for flash memory takes 1 to 25 seconds mainly depending on the flash capacity. Since the flash chip capacity is doubled in every year, this mounting time will soon become the most dominant reason of the delay of system start-up time. Therefore, in this paper, we present instant mounting techniques for flash file systems by storing the in-memory file system metadata to flash memory when unmounting the file system and reloading the stored metadata quickly when mounting the file system. These metadata snapshot techniques are specifically developed for NOR- and NAND-type flash memories, while at the same time, overcoming their physical constraints. The proposed techniques check the validity of the stored snapshot and use the proposed fast crash recovery techniques when the snapshot is invalid. Based on the experimental results, the proposed techniques can reduce the flash mounting time by about two orders of magnitude over the existing de facto standard flash file system.

A lifespan-aware reliability scheme for RAID-based flash storage

Due to the ever-growing capacity of flash memory along with its good properties such as low-power consumption and high performance, flash-based SSDs (solid state disks) are anticipated to be used in the storage of high-end server systems. However, the reliability problem of flash devices is becoming increasingly serious. The number of P/E (program/erase) cycles allowed to each flash block is too small, especially less than 10,000 for MLC (multi-level cell) flash memory. Furthermore, the bit error rate of flash memory becomes rapidly high as the number of P/E cycles increases. To relieve these problems, we present a lifespan-aware reliability scheme, which adopts RAID technologies together with ECCs (error correction codes). First, our scheme dynamically manages the size of striping group to cope with the increasing error rates of flash memory as the number of P/E cycles increases. Second, we use a device-aware log block mapping scheme, which uses different reliability policies for data blocks and log blocks by taking advantage of the characteristics of each block type. Third, we use small amount of storage class memory (SCM) to save parity blocks temporarily. By absorbing frequent updates of parity into SCM, we can extend the lifespan of flash memory. Simulation experiments show that our scheme obtains high reliability with minimum space overhead as well as improved I/O performances compared to traditional RAID-5.

FeGC: An efficient garbage collection scheme for flash memory based storage systems

NAND flash memory is a promising storage media that provides low-power consumption, high density, high performance, and shock resistance. Due to these versatile features, NAND flash memory is anticipated to be used as storage in enterprise-scale systems as well as small embedded devices. However, unlike traditional hard disks, flash memory should perform garbage collection that consists of a series of erase operations. The erase operation is time-consuming and it usually degrades the performance of storage systems seriously. Moreover, the number of erase operations allowed to each flash memory block is limited. This paper presents a new garbage collection scheme for flash memory based storage systems that focuses on reducing garbage collection overhead, and improving the endurance of flash memory. The scheme also reduces the energy consumption of storage systems significantly. Trace-driven simulations show that the proposed scheme performs better than various existing garbage collection schemes in terms of the garbage collection time, the number of erase operations, the energy consumption, and the endurance of flash memory.

A New Linux Swap System for Flash Memory Storage Devices

Flash memory has become an alternative storage device for future computer systems. Legacy operating system functionalities like file system and memory system should be modified since flash memory cannot be overwritten unless erased in advance. In this paper, we study a new Linux swap system considering flash memory as a swap storage. Especially we focus on the garbage collection performance and reduce the number of erasures and the number of data copies due to garbage collection. We collected real traces of swap I/O activities of Linux kernel and show that proposed method shows much better performance by means of trace-driven simulation.

Popular channel concentration schemes for efficient channel navigation in internet protocol televisions

Internet Protocol Television (IPTV) is becoming increasingly popular as an emerging Internet application due to its variety of contents and services. However, hundreds of IPTV channels make it difficult to find one¿s desired channel. To relieve this problem, this paper presents a popular channel concentration scheme for efficient channel navigation in IPTV. Based on the property that TV channel selections concentrate on several frequently selected channels, the proposed scheme reorganizes channel sequences by clustering popular channels. Specifically, channels are rearranged in a frequency interleaved way to minimize the seek distance. Simulation studies show that the proposed scheme reduces the seek time of IPTV channel navigation significantly when up-down channel selection interfaces are used.

LeCramFS: an efficient compressed file system for flash-based portable consumer devices

Cost-effectiveness is one of the most critical factors in the development of portable consumer devices. The use of compression techniques is a simple but effective solution for achieving such cost-effectiveness. This paper gives a comprehensive analysis of file system level compression techniques for portable consumer devices. Since conventional compressed file systems are optimized for disk like devices and relatively abundant computing resources, they are not suitable for portable devices with small amount of memory and weak processing power. This paper presents the design and implementation of LeCramFS, which is a new read-only compressed file system designed for small portable devices and flash memory. Experiments by implementation show that the proposed file system outperforms conventional ones in terms of memory-efficiency and I/O performance.

Replica-aware caching for Web proxies

A significant percentage of Web objects are replicas. For example, a vast majority of image files such as banners, buttons, and logos are duplicated throughout the WWW. Nevertheless, Web caching systems generally treat the replicas as different objects because they have different URLs. In this paper, we propose a simple and efficient way to manage the replicated objects for Web proxy caches. In the proposed scheme, the MD5 checksum, together with the size of an object, forms an identifier of a Web object that can distinguish replicas. Experimental results show that the proposed scheme significantly improves the cache hit rate and the byte hit rate by removing the redundant objects from the cache and reflecting the popularity of objects more precisely.

Hybrid transcoding for QoS adaptive video-on-demand services

Transcoding is a core technique that is used in providing quality-of-service (QoS) adaptive video-on-demand service. Many studies have examined how best is to perform transcoding and reduce computation overhead. However, the question of when to transcode has not been adequately studied in previous research. This paper addresses this question and presents a simple and intelligent approach that can be used to reduce both disk bandwidth and space requirements. Our approach determines the optimum time to apply transcoding by considering the potential benefits that can be realized. For instance, in order to save disk bandwidth for frequently accessed content, it pre-creates and stores multiple QoS versions. On the other hand, in order to save disk space for rarely accessed content, it stores only a single QoS version and performs transcoding on the fly. The key is to find the optimal threshold between pre-created multiple QoS versions and on-demand transcoding. We compute the optimal threshold by using a mathematical model. A simulation-based experiment to evaluate the effectiveness of our new approach highlights three advantages. First, our method effectively reduces both disk bandwidth and space requirements. Second, our technique is more efficient for skewed access patterns. Third, the threshold computed by our mathematical model results in improved performance regardless of environmental parameters.