• Journal of KIISE:Computer Systems and Theory
• /
• v.37 no.6
• /
• pp.330-338
• /
• 2010

Flash storage devices are very popularly used in portable devices such as cell phones, PDAs and MP3 players. As technology is improved, users want much bigger and faster storage system. Paradoxically, people have to wait more and more time proportionally to the capacity of their storage devices when these are trying to be recovered after file system crash. It is serious problem because booting time of devices is dominated by crash recovery of flash file system. In this paper, we design a crash recovery mechanism, named 'Working Area(WA hereafter)' technique, which has bounded crash recovery execution time. With WA technique, write operations to flash memory are only performed in WA. Therefore, by simply scanning the latest WA. We can recover a file system crash because every change for flash memory is occured only in latest WA. We implement the WA technique based on YAFFS2 and evaluate by comparing with traditional techniques. As a result, WA technique shows that its crash recovery execution time is 25 times faster than Log-based Method when we use 1 gig a bytes NAND flash memory in worst case. This gap will be futher and futher as storage capacity grows.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.11
• /
• pp.589-598
• /
• 2007

In the group communication which has multiple data streams and various access privileges, it is necessary to provide group access control. The group members having the same access privilege are classified into one class, and the classes form a hierarchy based on the access relations. Then each class is assigned to a secret key. In the previous schemes, a single logical key graph is constructed from the hierarchy and each member always holds all secret keys of the classes he can access in the proactive manner. Thus, higher-privileged members hold more keys then lower-privileged members. However, if the hierarchy is large, each member manages too many keys and the size of multicast message in rekeying increases in proportion to the size of the hierarchy. Moreover, most of the members access a small portion of multiple data streams simultaneously. Therefore, it is redundant to receive rekeying message and update the keys in which he is not currently interested. In this paper, we present a new key management scheme that takes a reactive approach in which each member obtains the key of a data stream only when he wants to access the stream. Each member holds and updates only the key of the class he belongs. If he wants to get the key of other class, he derives it from his key and the public parameter. Proposed scheme considerable reduces the costs for rekeying, especially in the group where access relations are very complex and the hierarchy is large. Moreover, the scheme has another advantage that it easily reflects the change of access relations.

• Journal of Intelligence and Information Systems
• /
• v.15 no.1
• /
• pp.123-140
• /
• 2009

Collaborative filtering recommends products using customers' preferences, so it cannot recommend products to the new customer who has no preference information. This paper proposes a novel approach to new customer recommendations using the social network analysis which is used to search relationships among social entities such as genetics network, traffic network, organization network, etc. The proposed recommendation method identifies customers most likely to be neighbors to the new customer using the centrality theory in social network analysis and recommends products those customers have liked in the past. The procedure of our method is divided into four phases : purchase similarity analysis, social network construction, centrality-based neighborhood formation, and recommendation generation. To evaluate the effectiveness of our approach, we have conducted several experiments using a data set from a department store in Korea. Our method was compared with the best-seller-based method that uses the best-seller list to generate recommendations for the new customer. The experimental results show that our approach significantly outperforms the best-seller-based method as measured by F1-measure.

• Journal of KIISE:Computer Systems and Theory
• /
• v.33 no.3
• /
• pp.178-192
• /
• 2006

The buffer overflow attack is the single most dominant and lethal form of security exploits as evidenced by recent worm outbreaks such as Code Red and SQL Stammer. In this paper, we propose microarchitectural techniques that can detect and recover from such malicious code attacks. The idea is that the buffer overflow attacks usually exhibit abnormal behaviors in the system. This kind of unusual signs can be easily detected by checking the safety of memory references at runtime, avoiding the potential data or control corruptions made by such attacks. Both the hardware cost and the performance penalty of enforcing the safety guards are negligible. In addition, we propose a more aggressive technique called corruption recovery buffer (CRB), which can further increase the level of security. Combined with the safety guards, the CRB can be used to save suspicious writes made by an attack and can restore the original architecture state before the attack. By performing detailed execution-driven simulations on the programs selected from SPEC CPU2000 benchmark, we evaluate the effectiveness of the proposed microarchitectural techniques. Experimental data shows that enforcing a single safety guard can reduce the number of system failures substantially by protecting the stack against return address corruptions made by the attacks. Furthermore, a small 1KB CRB can nullify additional data corruptions made by stack smashing attacks with only less than 2% performance penalty.

• Journal of KIISE:Computer Systems and Theory
• /
• v.28 no.12
• /
• pp.660-673
• /
• 2001

In Video-on-Demand (VoD) servers, disk throughput is an important system design parameter because it is directly related to the number of user requests that can be served simultaneously. In this paper, we propose an efficient periodic request grouping scheme for disk array-based VoD servers that reduces the disk seek time, thus improving the disk throughput of VoD disk arrays. To reduce the disk seek time, the proposed scheme groups the periodic requests that access data blocks stored in adjacent regions into one, and arranges these groups in a pre-determined order (e.g., in left-symmetric or right-symmetric fashion). Our simulation result shows that the proposed scheme reduces the average disk bandwidth required by a single video stream and can serve more user requests than existing schemes. For a data block size of 192KB, the number of simultaneously served user requests is increased by 8% while the average waiting time for a user request is decreased by 20%. We also propose an adaptation technique that conforms the proposed scheme to the user preference changes for video streams.

• Journal of KIISE:Computer Systems and Theory
• /
• v.26 no.11
• /
• pp.1318-1332
• /
• 1999

본 논문에서는 입출력 응용을 위해 collective I/O 기법을 기반으로 한 실행시간 시스템의 설계, 구현 그리고 그 성능평가를 기술한다. 여기서는 모든 프로세서가 동시에 I/O 요구에 따라 스케쥴링하며 I/O를 수행하는 collective I/O 방안과 프로세서들이 여러 그룹으로 묶이어, 다음 그룹이 데이터를 재배열하는 통신을 수행하는 동안 오직 한 그룹만이 동시에 I/O를 수행하는 pipelined collective I/O 등의 두 가지 설계방안을 살펴본다. Pipelined collective I/O의 전체 과정은 I/O 노드 충돌을 동적으로 줄이기 위해 파이프라인된다. 이상의 설계 부분에서는 동적으로 충돌 관리를 위한 지원을 제공한다. 본 논문에서는 다른 노드의 메모리 영역에 이미 존재하는 데이터를 재 사용하여 I/O 비용을 줄이기 위해 collective I/O 방안에서의 소프트웨어 캐슁 방안과 두 가지 모형에서의 chunking과 온라인 압축방안을 기술한다. 그리고 이상에서 기술한 방안들이 입출력을 위해 높은 성능을 보임을 기술하는데, 이 성능결과는 Intel Paragon과 ASCI/Red teraflops 기계 상에서 실험한 것이다. 그 결과 응용 레벨에서의 bandwidth는 peak point가 55%까지 측정되었다.Abstract In this paper we present the design, implementation and evaluation of a runtime system based on collective I/O techniques for irregular applications. We present two designs, namely, "Collective I/O" and "Pipelined Collective I/O". In the first scheme, all processors participate in the I/O simultaneously, making scheduling of I/O requests simpler but creating a possibility of contention at the I/O nodes. In the second approach, processors are grouped into several groups, so that only one group performs I/O simultaneously, while the next group performs communication to rearrange data, and this entire process is pipelined to reduce I/O node contention dynamically. In other words, the design provides support for dynamic contention management. Then we present a software caching method using collective I/O to reduce I/O cost by reusing data already present in the memory of other nodes. Finally, chunking and on-line compression mechanisms are included in both models. We demonstrate that we can obtain significantly high-performance for I/O above what has been possible so far. The performance results are presented on an Intel Paragon and on the ASCI/Red teraflops machine. Application level I/O bandwidth up to 55% of the peak is observed.he peak is observed.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.3_4
• /
• pp.145-151
• /
• 2004

We present an efficient data structure to obtain the range minima in an away in constant time. Recently, suffix ways are extensively used to search DNA sequences fast in bioinformatics. In constructing suffix arrays, solving the range minima problem is necessary When we construct suffix arrays, we should solve the range minima problem not only in a time-efficient way but also in a space-efficient way. The reason is that DNA sequences consist of millions or billions of bases. Until now, the most efficient data structure to find the range minima in an way in constant time is based on the method that converts the range minima problem in an array into the LCA (Lowest Common Ancestor) problem in a Cartesian tree and then converts the LCA problem into the range minima problem in a specific array. This data structure occupies O( n) space and is constructed in O(n) time. However since this data structure includes intermediate data structures required to convert the range minima problem in an array into other problems, it requires large space (=13n) and much time. Our data structure is based on the method that directly solves the range minima problem. Thus, our data structure requires small space (=5n) and less time in practice. As a matter of course, our data structure requires O(n) time and space theoretically.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.3_4
• /
• pp.224-238
• /
• 2004

Multi-modality image registration is a widely used image processing technique to obtain composite information from two different kinds of image sources. This study proposes an image registration method based on moment information and surface distance, which improves the previous surface-based registration method. The proposed method ensures stable registration results with low registration error without being subject to the initial position and direction of the object. In the preprocessing step, the surface points of the object are extracted, and then moment information is computed based on the surface points. Moment information is matched prior to fine registration based on the surface distance, in order to ensure stable registration results even when the initial positions and directions of the objects are very different. Moreover, surface comer sampling algorithm has been used in extracting representative surface points of the image to overcome the limits of the existed random sampling or systematic sampling methods. The proposed method has been applied to brain MRI(Magnetic Resonance Imaging) and PET(Positron Emission Tomography), and its accuracy and stability were verified through registration error ratio and visual inspection of the 2D/3D registration result images.

• Journal of KIISE:Computer Systems and Theory
• /
• v.27 no.1
• /
• pp.44-52
• /
• 2000

This paper is contented with packet transmission scheduling problem for repeating all-to-all broadcasts in WDM optical passive-star networks in which there are N nodes and k wavelengths. It is assumed that each node has one tunable transmitter and one fixed-tuned receiver, and each transmitter can tune to k different wavelengths. The tuning delay represents the time taken for a transmitter to tune from one wavelength to another and represented as ${\delta}$(>0) in units of packet durations. We define all-to-all broadcast as the one where every node transmits packets to all the other nodes except itself. So, there are in total N(N-1) packets to be transmitted for an all-to-all broadcast. The optimal transmission scheduling is to schedule In such a way that all packets can be transmitted within the minimum time. In this paper, we propose the condition for optimal transmission schedules and present an optimal transmission scheduling algorithm for arbitrary values of N, k, and ${\delta}$ The cycle length of the optimal schedules is $max{[\frac{N}{k}](M-1)$, $k{\delta}+N-1$}.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.5_6
• /
• pp.249-256
• /
• 2007

GPS(generalized processor sharing) is a fair scheduling scheme that guarantees fair distribution of resources in an instantaneous manner, while virtual clock pursues fairness in the sense of long-term. In this paper, we notice that the degree of memorylessness is the key difference of the two schemes, and propose a unified scheduling model that covers the whole spectrum of history-sensitiveness. In this model, each application's resource right is represented in a value called deposit, which is accumulated at a predefined rate and is consumed for services. The unused deposit, representing non-usage history, gives the application more opportunity to be scheduled, hence relatively enhancing its response time. Decay of the deposit means partial erase of the history and, by adjusting the decaying rate, the degree of history-sensitiveness is controlled. In the spectrum, the memoryless end corresponds GPS and the other end with full history corresponds virtual clock. And there exists a tradeoff between average delay and long-term fairness. We examine the properties of the model by analysis and simulation.