• Journal of KIISE:Databases
• /
• v.28 no.3
• /
• pp.384-394
• /
• 2001

In this paper, we propose AST-AET(Average valid Start Time-Average valid End Time) data migration strategy based on the storage structure where temporal data is divided into a past segment, a current segment, and a future segment. We define AST and AET which are used in AST-AET data migration strategy and also define entity versions to be stored in each segment. We describe methods to compute AST and AET, and processes to search entity versions for migration and move them. We compare average response times for user queries between AST-AET data migration strategy and the existing LST-GET(Least valid Start Time-Greatest valid End Time) data migration strategy. The experimental results show that, when there are no LLTs(Long Lived Tuples), there is little difference in performance between the two migration strategies because the size of a current segment is nearly equal. However, when there are LLTs, the average response time of AST-AET data migration strategy is smaller than that of LST-GET data migration strategy because the size of a current segment of LST-GET data migration strategy becomes larger. In addition, when we change average interarrival times of temporal queries, generally the average response time of AST-AET data migration strategy is smaller than that of LST-GET data migration strategy.

• Journal of KIISE:Databases
• /
• v.28 no.3
• /
• pp.512-527
• /
• 2001

In a client-server spatial database, it is desirable to maintain the cached data in a client side to minimize the communication overhead across a network. This paper deals with the issues of concurrency and consistency of map updates in this environment. A client transaction to update map data is an interactive work and takes a long time to complete it. The map update in a client site may affect the other sites'updates because of dependencies between spatial data stored at different sites. The concurrent updates should be propagated to the other clients as well as the server to keep the consistency of map replicated in a client cache, and also the communication overhead of the update propagation should be minimized not to lose the benefit of caching. The newly proposed cache region locking with CR lock and CX lock controls the update dependency due to spatial relationships. CS lock and COD lock are suggested to use optimistic detection-based approaches for guaranteeing the consistency of cached client data. The cooperative update protocol uses these extended locking primitives and Spatial Relationship-based 2PC (SR-based 2PC). This paper argues that the concurrent updates of cached client spatial data can be achieved by deciding on collaborative updates or independent updates based on spatial relationships.

• Journal of KIISE:Databases
• /
• v.30 no.5
• /
• pp.451-463
• /
• 2003

Recently, there have been various research efforts to develop strategies for accelerating OLAP operations on huge amounts of spatio-temporal data. Most of the work is based on multi-tree structures which consist of a single R-tree variant for spatial dimension and numerous B-trees for temporal dimension. The multi~tree based frameworks, however, are hardly applicable to spatio-temporal OLAP in practice, due mainly to high management cost and low query efficiency. To overcome the limitations of such multi-tree based frameworks, we propose a new approach called Hilbert Cube(H-Cube), which employs fractals in order to impose a total-order on cells. In addition, the H-Cube takes advantage of the traditional Prefix-sum approach to improve Query efficiency significantly. The H-Cube partitions an embedding space into a set of cells which are clustered on disk by Hilbert ordering, and then composes a cube by arranging the grid cells in a chronological order. The H-Cube refines cells adaptively to handle regional data skew, which may change its locations over time. The H-Cube is an adaptive, total-ordered and prefix-summed cube for spatio-temporal data warehouses. Our approach focuses on indexing dynamic point objects in static spatial dimensions. Through the extensive performance studies, we observed that The H-Cube consumed at most 20% of the space required by multi-tree based frameworks, and achieved higher query performance compared with multi-tree structures.

• Journal of KIISE:Databases
• /
• v.30 no.5
• /
• pp.507-520
• /
• 2003

In this paper, an automatic construction method of theme melody index for large music database and an associative content-based music retrieval mechanism in which the constructed theme melody index is mainly used to improve the users' response time are proposed. First, the system automatically extracted the theme melody from a music file by the graphical clustering algorithm based on the similarities between motifs of the music. To place an extracted theme melody into the metric space of M-tree, we chose the average length variation and the average pitch variation of the theme melody as the major features. Moreover, we added the pitch signature and length signature which summarize the pitch variation pattern and the length variation pattern of a theme melody, respectively, to increase the precision of retrieval results. We also proposed the associative content-based music retrieval mechanism in which the k-nearest neighborhood searching and the range searching algorithms of M-tree are used to select the similar melodies to user's query melody from the theme melody index. To improve the users' satisfaction, the proposed retrieval mechanism includes ranking and user's relevance feedback functions. Also, we implemented the proposed mechanisms as the essential components of content-based music retrieval systems to verify the usefulness.

• Journal of KIISE:Databases
• /
• v.28 no.3
• /
• pp.357-372
• /
• 2001

In this paper, we present the concept of generalization in constructing windows for subsequence matching and propose a new subsequence matching method. GeneralMatch, based on the generalization. The earlier work of Faloutsos et al.(FRM in short) causes a lot of false alarms due to lack of the point-filtering effect. DualMatch, which has been proposed by the authors, improves performance significantly over FRM by exploiting the point filtering effect, but it has the problem of having a smaller maximum window size (half that FRM) given the minimum query length. GeneralMatch, an improvement of DualMatch, offers advantages of both methods: it can use large windows like FRM and, at the same time, can exploit the point-filtering effect like DualMatch. GeneralMatch divides data sequences into J-sliding windows (generalized sliding windows) and the query sequence into J-disjoint windows (generalized disjoint windows). We formally prove that our GeneralMatch is correct, i.e., it incurs no false dismissal. We also prove that, given the minimum query length, there is a maximum bound of the window size to guarantee correctness of GeneralMatch. We then propose a method of determining the value of J that minimizes the number of page accesses, Experimental results for real stock data show that, for low selectivities ($10^{-6}~10^{-4}$), GeneralMatch improves performance by 114% over DualMatch and by 998% iver FRM on the average; for high selectivities ($10^{-6}~10^{-4}$), by 46% over DualMatch and by 65% over FRM on the average.

• Journal of KIISE:Databases
• /
• v.28 no.3
• /
• pp.279-300
• /
• 2001

Data mining plays an important role in knowledge discovery process and usually various existing algorithms are selected for the specific purpose of the mining. Currently, data mining techniques are actively to the statistics, business, electronic commerce, biology, and medical area and currently numerous algorithms are being researched and developed for these applications. However, in a long run, only a few algorithms, which are well-suited to specific applications with excellent performance in large database, will survive. So it is reasonable to focus our effort on those selected algorithms in the future. This paper classifies about 30 existing algorithms into 7 categories - association rule, clustering, neural network, decision tree, genetic algorithm, memory-based reasoning, and bayesian network. First of all, this work analyzes systematic hierarchy and characteristics of algorithms and we present 14 criteria for classifying the algorithms and the results based on this criteria. Finally, we propose the best algorithms among some comparable algorithms with different features and performances. The result of this paper can be used as a guideline for data mining researches as well as field applications of data mining.

• Journal of KIISE:Databases
• /
• v.28 no.3
• /
• pp.494-511
• /
• 2001

In the environment that database management systems manage disk spaces for storing databases directly, this paper proposes a correct and efficient scheme of performing pending actions for the removal of database files. As for performing pending actions, upon performing recovery, the recovery process must identify unperformed pending actions of not-yet-terminated transactions and then perform those actions completely. Making the recovery process identify those actions through the analysis of log records in the log file is the basic idea of this paper. This scheme, as an extension of the execution of transactions, fuzzy checkpoint, and recovery of ARIES, uses the following methods: First, to identify not-yet-terminated transactions during recovery, transactions perform pending actions after writing 'pa_start'log records that signify both the commit of transactions and the start of executing pending actions, and then write 'eng'log records. Second, to restore pending-actions-lists of not-yet-terminated transactions during recovery, each transaction records its pending-actions-list in 'pa_start'log record and the checkpoint process records pending-actions-lists of transactions that are decided to be committed in 'end_chkpt'log record. Third, to identify the next pending action to perform during recovery, whenever a page is updated during the execution of pending actions, transactions record the information that identifies the next pending action to perform in the log record that has the redo information against the page.

• Journal of KIISE:Databases
• /
• v.30 no.5
• /
• pp.438-450
• /
• 2003

Spatial joins find pairs of objects that overlap with each other. In spatial joins using indexes, original-space indexes such as the R-tree are widely used. An original-space index is the one that indexes objects as represented in the original space. Since original-space indexes deal with sizes of objects, it is difficult to develop a formal algorithm without relying on heuristics. On the other hand, transform-space indexes, which transform objects in the original space into points in the transform space and index them, deal only with points but no sites. Thus, spatial join algorithms using these indexes are relatively simple and can be formally developed. However, the disadvantage of transform-space join algorithms is that they cannot be applied to original-space indexes such as the R-tree containing original-space objects. In this paper, we present a novel mechanism for achieving the best of these two types of algorithms. Specifically, we propose a new notion of the transform-space view and present the transform-space view join algorithm(TSVJ). A transform-space view is a virtual transform-space index based on an original-space index. It allows us to interpret on-the-fly a pre-built original-space index as a transform-space index without incurring any overhead and without actually modifying the structure of the original-space index or changing object representation. The experimental result shows that, compared to existing spatial join algorithms that use R-trees in the original space, the TSVJ improves the number of disk accesses by up to 43.1% The most important contribution of this paper is to show that we can use original-space indexes, such as the R-tree, in the transform space by interpreting them through the notion of the transform-space view. We believe that this new notion provides a framework for developing various new spatial query processing algorithms in the transform space.

• The KIPS Transactions:PartC
• /
• v.16C no.3
• /
• pp.285-298
• /
• 2009

According to the development of ubiquitous technologies, sensor networks is used in various area. In particular, medical field is one of the significant application areas using sensor networks, and recently it has come to be more important according to standardization of the body sensor networks technology. There are special characteristics of their own for medical sensor networks, which are different from the one of sensor networks for general application or environment. In this paper, we propose a hierarchical medical sensor networks structure considering own properties of medical applications, and also introduce transmission mechanism based on hierarchical structure. Our mechanism uses the priority and threshold value for medical sensor nodes considering patient's needs and health condition. Through this way Cluster head can transmit emergency data to the Base station rapidly. We also present the new key establishment mechanism based on key management mechanism which is proposed by L. Eschenauer and V. Gligor for our proposed structure and transmission mechanism. We use key provisioning for emergency nodes that have high priority based on patients' health condition. This mechanism guarantees the emergency nodes to establish the key and transmit the urgent message to the new cluster head more rapidly through preparing key establishment with key provisioning. We analyze the efficiency of our mechanism through comparing the amount of traffic and energy consumption with analysis and simulation with QualNet simulator. We also implemented our key management mechanism on TmoteSKY sensor board using TinyOS 2.0 and through this experiments we proved that the new mechanism could be actually utilized in network design.

• KIPS Transactions on Computer and Communication Systems
• /
• v.8 no.12
• /
• pp.287-296
• /
• 2019

Due to the development and increased usage of Internet services such as IoT and cloud computing, a large number of packets are being generated on the Internet. In order to create a safe Internet environment, malicious data that may exist among these packets must be processed and detected quickly. In this paper, we apply MongoDB, which is specialized for unstructured data analysis and big data processing, to intrusion detection system for rapid processing of big data security events. In addition, building the intrusion detection system(IDS) using some of the private cloud resources which is the target of protection, elastic and dynamic reconfiguration of the IDS is made possible as the number of security events increase or decrease. In order to evaluate the performance of MongoDB - based IDS proposed in this paper, we constructed prototype systems of IDS based on MongoDB as well as existing relational database, and compared their performance. Moreover, the number of virtual machine has been increased to find out the performance change as the IDS is distributed. As a result, it is shown that the performance is improved as the number of virtual machine is increased to make IDS distributed in MongoDB environment but keeping the overall system performance unchanged. The security event input rate based on distributed MongoDB was faster as much as 60%, and distributed MongoDB-based intrusion detection rate was faster up to 100% comparing to the IDS based on relational database.