| 研究課題/領域番号 |
14J07114
|
|---|
| 研究機関 | 会津大学 |
|---|
研究代表者 |
陳 武輝 会津大学, コンピュータ理工学研究科, 特別研究員(PD)
|
|---|
| 研究期間 (年度) |
2014-04-25 – 2016-03-31
|
| キーワード | Situation awareness / Mapreduce / data placement / big data infrastructure |
|---|
| 研究実績の概要 |
I summarized the main tasks which have been done as follows:
First, an active situation awareness (SA) framework for social network services has been developed. Active situation awareness for the cases of social network services (SNS) was illustrated. Perception by mining SNS data using TF-IDF, comprehension of the information by inference of ontology and rule, projection of new suggest by rule inference were explained. Finally system architecture and evaluation of the system was done. Our new framework shows more enhanced situation awareness system and new vision of awareness computing.
Second, a Big Data Infrastructure for extraction of perception information based on the SA framework has been built. I have constructed a real cluster testbed of 18 data nodes and ran MapReduce jobs on input data that are distributed in it using Hadoop 2.2.0. The testbed consisted of three subclusters, each with three racks and six data nodes.
Third, I have developed a novel optimal data placement technique to improve the performance of big data infrastructure for data mining algorithm on perception layer. I first formulated the problem of optimal data placement and proposed a generative model to minimize global data access costs in data centers. Then, I presented the theoretical analysis of RBT and RST, and proposed a topology-aware heuristic algorithm to construct an optimal RDT based on them. The experimental results demonstrated that our optimal data replacement approach could minimize global data access cost effectively with low communication costs and less computation costs.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
First, two important parts in this research plan has been done in first year as follows:
1) I have constructed a Hadoop system with 18 data nodes to be able to simulate an environment where nodes are distributed on geographically. Based on the environment built, I have developed a novel heuristic algorithm to allocate data blocks optimally for geo-aware Big data computation.
2) I have constructed a general active situation awareness system with three layered awareness architecture with lab members. The system can understand a new situation or projection from social network services or Web data service. Web APIs for providing new facts to be used in comprehension engine by rule inference have been developed. Our architecture shows a new framework to provide active situation awareness on SNS services.
Second, the following part can be finished soon: I am developing a data-placement technique for data-center resizing that considers both the static and dynamic characteristics of data centers. The objective is to find an optimal data placement to ameliorate the performance degradation when turning off idle nodes, and to improve the performance of big-data processing when turning on more working nodes, by optimizing the global data-transfer cost. some initial and positive results of this research have been observed.
|
| 今後の研究の推進方策 |
First, finish the following task which is in work-in-progress status: construction of Services-Of-A-Service (SOAS) model on Big Data and social links patterns, GSSN with scale free network on Big Data infrastructure, exploiting Map-Reduce approach on GSSN to locate and compose services, and evaluating performance of discovery and composition on GSSN with Hadoop. SOAS model will be constructed on linked service model, property model for link quality and recommendation algorithm for GSSN, optimal exploitation algorithm with Map-Reduce operation on the graph and whole implementation and evaluation will be carried out.
Second, in the coming year, I will construct a novel situation awareness framework combining GSSN to perceive, comprehend, and project new situations about plans for power saving at now and prevention of difficulties or problems from earthquake or other disasters possible in the future.
|
