Myeong-Cheol Ko

CPOC: effective static task scheduling for grid computing

Effective task scheduling is crucial for achieving good performance in high performance computing. Many scheduling algorithms have been devised for heterogeneous computing and CPOP is one of the scheduling algorithms. In this paper we present new scheduling algorithms, CPOC and CPOC_E by modifying the CPOP. We use a cluster of processors for critical-path tasks while a single processor is used in the CPOP. This heuristic is useful for realistic Grid computing environments in which communication costs are not arbitrarily heterogeneous. In an additional heuristic the critical-path tasks are considered to finish (or start) as early as possible when non critical-path tasks are scheduled. For performance study we developed a task graph generator and a tool which would support more realistic network configuration. The experimental results show our scheduling algorithm outperforms the CPOP as well as the HEFT.

Bitmap-based Prefix Caching for Fast IP Lookup

IP address lookup is very crucial in performance of routers. Several works have been done on prefix caching to enhance the performance of IP address lookup. Since a prefix represents a range of IP addresses, a prefix cache shows better performance than an IP address cache. However, not every prefix is cacheable in itself. In a prefix cache it causes false hit to cache a non-leaf prefix because there is possibly the longer matching prefix in the routing table. Prefix expansion techniques such as complete prefix tree expansion (CPTE) make it possible to cache the non-leaf prefixes as the expanded forms, but it is hard to manage the expanded prefixes. The expanded prefixes sometimes incur a great deal of update overhead in a routing table. We propose a bitmap-based prefix cache (BMCache) to provide low update overhead as well as low cache miss ratio. The proposed scheme does not have any expanded prefixes in the routing table, but it can expand a non-leaf prefix using a bitmap on caching time. The trace-driven simulation shows that BMCache has very low miss ratio in spite of its low update overhead compared to other schemes.

Ontology based integration of web databases by utilizing web interfaces

The amount of information which we can get from web environments, especially in the web databases, has rapidly grown. And, many questions often can be answered using integrated information than using single web database. Therefore the need for integrating web databases became increasingly. In order to make multiple web databases to interoperate effectively, the integrated system must know where to find the relevant information which is concerned with a users query on the web databases and which entities in the searched web databases meet the semantics in the users query. To solve these problems, this paper presents an approach which provides ontology based integration method for multiple web databases. The proposed approach uses the web interfaces through which generally a user can acquire desired information in the web databases.

A Term Cluster Query Expansion Model Based on Classification Information in Natural Language Information Retrieval

A natural language information retrieval system ranks related documents according to criteria based on user query keywords and document similarities. However, many efforts have been made to make more useful query keywords because users do not use many keywords in their natural language search query when retrieving information on the Web. Because a keyword does not provide much information, however, relevance feedback is generally used to complement the weakness of general retrieval methods. This paper proposes a term cluster query expansion model based on classification information of retrieved documents. This model generates classification information from the upper ranked n documents retrieved by retrieval system. On the basis of the extracted classification information, the term cluster (m) that represents each group is generated, and then the model allows user to select term cluster that corresponds to user information needs. The query keywords are expanded by using a relevance feedback algorithm based on the selected classification information. As a result of the experiments with test collection, the retrieval effectiveness was improved by 13.2% compared to the initial query when the Rocchio method was used.

A Resolving of Word Sense Ambiguity Using Two-level Document Ranking Method in Information Retrieval

The resolving of word sense ambiguity is one of important problem in information retrieval. Words can have several meanings or senses, and a use of the word out of context is ambiguous between these senses. Mutual information is a relation measure which represents relation between a word and another word. To disambiguate the lexical ambiguity in the retrieved documents, we re-evaluate relation between terms of retrieved document and terms of query. In this paper, we discuss a resolving of word sense ambiguity of information retrieval that is two-level document ranking method using mutual information. At the first level, we retrieve documents based on an automatically constructed keyword index. At the second-level, we reorder the retrieved documents using mutual information. We show that our method provides a resolving of word sense ambiguity in the retrieved documents. Also, we analyze several formulas and found a formula which to disambiguate the word sense ambiguity in the retrieved documents.

The roles of ontology and metadata registry for interoperable databases

In order to make multiple autonomous databases to interoperate effectively, semantic heterogeneities have to be detected and resolved Another difficulty is that users can be allowed to handle information easily from different heterogeneous databases that refer to the same real-world entity To solve these problems, in this paper, I present an information integration system for interoperable databases using metadata registry and ontology A metadata registry is a place to keep facts about characteristics of data that are necessary for data sharing and exchange in a specific domain An ontology defines concepts and relations among concepts The purpose of the proposed architecture is to define an information integration model, which combines characteristics of both standard specification of metadata registry and functionality of ontology for the concepts and relations.

A Hybrid IP Forwarding Engine with High Performance and Low Power

Many IP forwarding engines have employed TCAM for lookup operations due to its capability of storing variable-sized network prefixes. Power consumption has been one of the most important issues in TCAM-based engines. In contrast, SRAM can be operated with low power and latency but it cannot store variable-sized data directly. While it may need several accesses per lookup, TCAM needs only one access because of its parallel search. In this paper we propose a hybrid IP forwarding engine which elaborately combines TCAM and SRAM to allow both low power and high throughput. It consists of three stages based on different kinds of memories. Each stage may or may not be operated depending on a given IP address to maximize performance and save energy. Experiment results show that the proposed engine is at least 7.3 times faster than the normal TCAM-based engine with only 1.8% energy of that engine.

Effective static task scheduling for realistic heterogeneous environment

Effective task scheduling is crucial for achieving good performance in high performance computing. Many scheduling algorithms have been devised for heterogeneous computing, but most of algorithms have not been considered in realistic heterogeneous environments which are not arbitrarily heterogeneous but have locality in communication. In this paper we present new scheduling algorithms by considering the locality. It is thought that critical-path tasks are often important in reducing schedule length, however one of the previous scheduling algorithms, CPOP (Critical-Path-On-a-Processor) does not show good result against to expectation. Our first heuristic uses a cluster of processors for critical-path tasks while a single processor is used in the CPOP. This heuristic well exploits realistic computing environments in which communication costs are not arbitrarily heterogeneous. In an additional heuristic the critical-path tasks are considered to finish (or start) as early as possible when even non critical-path tasks are scheduled. For a performance study five scheduling algorithms are compared by experimenting on three different environments. The experimental results show our scheduling algorithm outperforms the others in the realistic heterogeneous environments.

Error metric for perceptual features preservation in polygonal surface simplification

Polygonal surface models are generally composed of a large amount of polygonal patches. Highly complex models can provide a convincing level of realism but rather cause problems in real-time applications, such as virtual reality or 3D simulation system focusing on real-time interactivity. Therefore, it is useful to have various simplified versions of the model according to the performance of the system. In this paper, we present a surface simplification algorithm, which can excellently preserve the characteristic features of the original model, even after drastic simplification process. In addition, the proposed algorithm is efficient in memory usage and useful in real-time rendering applications requiring progressive and incremental rendering of surface data.

Development of the Interactive Conference System in Ubiquitous Computing Environment

The lecture scenario of the previous works implicates fundamental didactic problems due to the lack of interactivity and opportunity for feedback. Current ubiquitous computing environment on the campus provides a good chance to apply the advanced teaching paradigm to the university lecture. In this paper, we develop the conference-oriented lecture system, which all attendees can interactively participate in the learning processes. The whole procedures of the lecture are automatically processed in the ubiquitous computing environment, so we can minimize the unnecessary time-consuming procedures.