• KIPS Transactions on Computer and Communication Systems
• /
• v.3 no.12
• /
• pp.433-442
• /
• 2014

In this paper, we introduce the client side cache of distributed file system for enhancing read performance by eliminating the network latency and decreasing the back-end storage burden. This performance enhancement can expand the fields of distributed file system to not only cloud storage service but also high performance storage service. This paper shows that the distributed file system with client side SSD cache can satisfy the requirements of VDI(Virtual Desktop Infrastructure) storage. The experimental results show that full-clone is more than 2 times faster and boot time is more than 3 times faster than NFS.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.15 no.6
• /
• pp.129-134
• /
• 2015

This paper analyzes file access characteristics of smartphone apps and finds that a large portion of file writes are performed only once. Based on this observation, we present a new buffer cache management scheme that considers this characteristics. Buffer cache improves storage performance by maintaining hot file data in memory thereby servicing subsequent requests without storage accesses. However, it should flush modified data to storage in order to resist system crashes. The proposed scheme evicts cache data that has been written only once upon flushes, thus improving cache space utilization. Simulation experiments show that the proposed scheme improves cache hit ratio by 5-33% and power consumption by 27-92%.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.2
• /
• pp.1-8
• /
• 2013

TIn this paper, we propose a high performance L1 cache structure on the high clock CPU of 4GHz. The proposed cache memory consists of three parts, i.e., a direct-mapped cache to support fast access time, a two-way set associative buffer to exploit temporal locality, and a buffer-select table. The most recently accessed data is stored in the direct-mapped cache. If a data has a high probability of a repeated reference, when the data is replaced from the direct-mapped cache, the data is selectively stored into the two-way set associative buffer. For the high performance and low power consumption, we propose an one way among two ways set associative buffer is selectively accessed based on the buffer-select table(BST). According to simulation results, Energy $^*$ Delay product can improve about 45%, 70% and 75% compared with a direct mapped cache, a four-way set associative cache, and a victim cache with two times more space respectively.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.2
• /
• pp.27-32
• /
• 2021

This paper proposes an energy-efficient cache and memory management scheme for next-generation IoT devices. The proposed scheme adopts a low-power phase-change memory (PCM) as the main memory of IoT devices, aims at minimizing the write traffic to PCM, which is vulnerable to write operations. Specifically, when a cache block of the last-level cache memory is flushed to main memory, the cache block that causes less writes to PCM is preferentially replaced by tracking the modifications of each cache line that constitutes the cache block. In addition, by considering the reference bit of the cache block and the dirty bit of the cache lines, our scheme reduces the energy consumption without degrading the memory system performances. Through simulations using SPEC benchmarks, it is shown that the proposed scheme reduces the write traffic to PCM by 34.6% on average and the power consumption by 28.9%, without memory performance degradations.

• Proceedings of the Korean Information Science Society Conference
• /
• 2008.06b
• /
• pp.366-370
• /
• 2008

NAND 플래시 메모리는 특성상 쓰기 횟수가 제한적이라는 단점을 가지고 있어 쓰기 연산이 빈번히 발생하게 되면 NAND 플래시 메모리의 수명이 줄어든다. 이러한 문제점을 해결하기 위해 NAND 플래시 메모리의 특성을 고려한 지연 쓰기 기법이 연구되고 있다. 하지만 지연 쓰기를 하기 때문에 쓰기 횟수는 줄어들지만 캐시 적중률이 낮아진다. 이러한 문제해결을 위해 본 논문에서는 NAND 플래시 메모리 기반 파일 시스템을 위한 더블 캐시 정책을 제안한다. 더블 캐시는 실질적인 캐시인 Real Cache와 요구 페이지의 패턴을 관찰하기 위한 Ghost Cache로 구성된다. 이 정책은 Real Cache에서의 지연 쓰기를 하지 않고, Ghost Cache 공간에서 dirty페이지와 clean페이지를 활용하여 효율적인 지연 쓰기가 가능하도록 설계함으로써 쓰기 횟수를 줄이고, 적중률을 높인다.

• International Journal of Contents
• /
• v.3 no.1
• /
• pp.19-23
• /
• 2007

Recently, several main-memory index structures have been proposed to reduce the impact of secondary cache misses. In mainmemory storage systems, secondary cache misses have a substantial effect on the performance of index structures. However, recent studies still stiffer from secondary cache misses when visiting each level of index tree. In this paper, we propose a new index structure that minimizes the total amount of cache miss latency. The proposed index structure prefetched grandchildren of a current node. The basic structure of the proposed index structure is based on that of the CSB+-Tree, which uses the concept of a node group to increase fan-out. However, the insert algorithm of the proposed index structure significantly reduces the cost of a split. The superiority of our algorithm is shown through performance evaluation.

• IEIE Transactions on Smart Processing and Computing
• /
• v.4 no.2
• /
• pp.78-82
• /
• 2015

General-purpose graphics processing units (GPGPUs) provide tremendous computational and processing power. Despite the latency hiding mechanism, a GPU architecture requires high memory bandwidth and lower latency between computational units and the memory system. For this reason, the current GPU architecture has private L1 caches in each core and a shared L2 cache to increase performance by reducing memory latency. But in some cases, this CPU-like cache design is not suitable for GPGPUs. In this paper, we analyze detailed cache performance related to GPGPU application characteristics, and suggest technical alternatives for the GPGPU architecture as future work.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.4
• /
• pp.108-116
• /
• 2013

Many algorithms have been introduced to improve performance by using massively parallel systems, which consist of several hundreds of processors. A typical example is a GPU system of many processors which uses shared memory. In the case of image filtering algorithms, which make references to neighboring points, the shared memory helps improve performance by frequently accessing adjacent pixels. However, using shared memory requires rewriting the existing codes and consequently results in complexity of the codes. Recent GPU systems support both L1 and L2 cache along with shared memory. Since the L1 cache memory is located in the same area as the shared memory, the improvement of performance is predictable by using the cache memory. In this paper, the performance of cache and shared memory were compared. In conclusion, the performance of cache-based algorithm is very similar to the one of shared memory. The complexity of the code appearing in a shared memory system, however, is resolved with the cache-based algorithm.

• Journal of Communications and Networks
• /
• v.14 no.3
• /
• pp.286-295
• /
• 2012

To meet the increased demand of video-on-demand (VoD) services, peer-to-peer (P2P) mesh-based multiple video approaches have been recently proposed, where each peer is able to find a video segment interested without resort to the video server. However, they have not considered the constraint of the server's upload bandwidth and the fairness between upload and download amounts at each peer. In this paper, we propose a novel P2P VoD streaming system, named peer cache adaptation (PECAN) where each peer adjusts its cache capacity adaptively to meet the server's upload bandwidth constraint and achieve the fairness. For doing so, we first propose a new cache replacement algorithm that designs the number of caches for a segment to be proportional to its popularity. Second, we mathematically prove that if the cache capacity of a peer is proportional to its segment request rate, the fairness between upload and download amounts at each peer can be achieved. Third, we propose a method that determines each peer's cache capacity adaptively according to the constraint of the server's upload bandwidth. Against the proposed design objective, some selfish peers may not follow our protocol to increase their payoff. To detect such peers, we design a simple distributed reputation and monitoring system. Through simulations, we show that PECAN meets the server upload bandwidth constraint, and achieves the fairness well at each peer. We finally verify that the control overhead in PECAN caused by the search, reputation, and monitoring systems is very small, which is an important factor for real deployment.

• Journal of KIISE
• /
• v.42 no.2
• /
• pp.177-182
• /
• 2015

The era of multi-core processors has begun since the limit of the clock speed has been reached. These days, multi-core technology is used not only in desktops, servers, and table PCs, but also in smartphones. In this architecture, there is always interference between processes, because of the sharing of system resources. To address this problem, cache partitioning is used, which can be roughly divided into two types: software and hardware cache partitioning. When it comes to dynamic cache partitioning, hardware cache partitioning is superior to software cache partitioning, because it needs no page copy. In this paper, we compare the effectiveness of hardware and software cache partitioning on the AMD Opteron 6282 SE, which is the only commodity processor providing hardware cache partitioning, to see whether this technique can be effectively deployed in dynamic environments.