Yasusi Kanada

A method of geographical name extraction from Japanese text for thematic geographical search

A text retrieval method called the thematic geographical search method has been developed and applied to a Japanese encyclopedia called the World Encyclopædia. In this method, the user specifies a search theme using free words, then obtains a sorted list of excerpts and hyperlinks to encyclopedia sentences that contain geographical names. Using this list, the user can also open maps that indicate the locations of the names. To generate an index of names for this searching, a method of extracting geographical names has been developed. In this method, geographical names are extracted, matched to names in a geographical name database, and identified. Geographical names, however, often have several types of ambiguities. Ambiguities are resolved by using non-local context analysis, which uses a stack and several other techniques. As a result, the precision of extracted names is more than 96% on average. This method depends on features of the Japanese language, but the strategy and most of the techniques can be applied to texts in English or other languages.

Network-virtualization nodes that support mutually independent development and evolution of node components

“Virtualization nodes” (VNodes) for programmable network-virtualization platforms are being developed. Criteria for “clean” network-virtualization are devised and applied to this platform and slices (virtual networks). These criteria meet one of the challenges targeted by the Virtualization Node Project, that is, to enable mutually independent development and evolution of components (namely, computational components called programmers and networking components called redirectors) in VNodes. To meet this challenge, the redirector plays the central role in implementing the following two functions of VNodes. The first function is creation of mapping between virtual links to external physical paths and mapping between virtual links to internal physical paths, which makes it possible to hide various alternative computational components in the VNode from the external network and to hide these external-network representations from the internal components. The second function is implementation of high-performance data conversion, which connects the external and internal data formats or mappings, by using an add-on card with a network processor. Two results are obtained from the performance evaluation of these functions. First, the overhead caused by mapping creation can be hidden by other tasks under normal conditions, but the overhead caused by mapping deletion must be reduced. Second, the data-conversion rate is half the wire rate, which should be increased in future work.

Taxonomy and Description of Policy Combination Methods

To control complicated and decomposable networking functions, such as Diffserv, two or more policies must cooperate. Combining two or more mutually dependent policies for a specific purpose is called policy combination. Methods of passing information between combined policies can be classified into real tags and virtual tags, or labels and attributes. Policy combinations can be classified into concatenation, parallel application, selection, and repetition. Explicitly specifying policy combinations makes policy systems semantically clearer and better suited to general use, extends the range of functionality, and improves the possibility of optimization. If policy combinations can be specified in a policy system, two types of policy organizations can be distinguished: homogeneous and heterogeneous. Heterogeneous organization is more service-oriented and seems to meet service-management requirements, but homogeneous organization is more device-oriented and may provide better performance.

A vectorization technique of hashing and its application to several sorting algorithms

A vectorized algorithm for entering data into a hash table is presented. A program that enters multiple data could not be executed on vector processors by conventional vectorization techniques because of data dependences. The proposed method enables execution of multiple data entry by conventional vector processors and improves the performance by a factor of 12.7, compared with the normal sequential method, when 4099 pieces of data are entered on the Hitachi S-810. This method is applied to address calculation sorting and the distribution counting sort, whose main part was unvectorizable by previous techniques. It improves performance by a factor of 12.8 when n=2/sup 14/ on the S-810.<<ETX>>

A representation of network node QoS control policies using rule-based building blocks

Network node functions, such as QoS or the security functions of routers, are becoming increasingly complex, so programs, not only configuration parameters, are required to control network nodes. In a policy-based network, a policy is defined at a policy server as a set of rules that deploy at network nodes where it must be translated into an executable program or parameters. Thus, a policy must be represented by a form in which the syntax and semantics are clearly defined, and which can be mechanically translated into an executable program. This is possible if the policy is written in an appropriate rule-based programming language. This paper describes such a language in which functions required for DiffServ can be specified for the interface between a policy server and network nodes. In this language, a policy rule can be composed using predefined primitive building blocks and control structures.

Accelerating non-numerical processing by an extended vector processor

The problems of utilizing an extended vector processor called IDP (Integrated Database Processor) for nonnumerical processing are discussed. IDP has highly pipelined merge and search vector facilities in addition to conventional vector facilities. This extended vector processor can be utilized not only for primitive relational operations but also for the following applications: (1) complex relational query using multiple indexes; (2) multiway merge/sort utility, and (3) generation and test processing. Vectorization methods and performance improvements are described and analyzed.<<ETX>>

A “network-paging” based method for wide-area live-migration of VMs

In cloud-computing environments, migration of virtual machines (VMs) between data centers can solve many problems such as load balancing and power saving. One of the difficulties in wide-area migration, however, is the “address-warping” problem, in which the address of the VM warps from the source server to the destination server. This confuses or complicates the status of the WAN, and the LANs connected to the WAN. We propose two solutions to this problem. One is to switch an address-translation rule, and the other is to switch multiple virtual networks. The former is analogous to paging in memory virtualization, and the latter is analogous to segmentation. The “network-paging” based method is described and our evaluation results are shown. It took less than 100 ms in average to switch from the source to the destination server using this method.

Two Rule-Based Building-Block Architectures for Policy-Based Network Control

Policy-based networks can be customized by users by injecting programs called policies into the network nodes. So if general-purpose functions can be specified in a policy-based network, the network can be regarded as an active network in the wider sense. In a policy-based network, two or more policies must often cooperate to provide a high-level function or policy. To support such building-block policies, two architectures for modeling a set of policies have been developed: pipe-connection architecture and label-connection architecture. It is shown that rule-based building blocks are better for policy-based network control and that the label-connection architecture is currently better. However, the pipe-connection architecture is better in regards to parallelism, which is very important in network environments.

Vectorization techniques for prolog

Several techniques for running Prolog programs on pipelined vector processors, such as the Hitachi S-820 or the Cray-2, are developed. This paper presents an automatic program transformation (vectorization) method of Prolog, which enables a type of or-parallel execution of Prolog programs using vector operations. Performance is evaluated on the Hitachi S-810 using the Eight-Queens problem. Its vector execution speed is 4.5 MLIPS (18 ms). This is eight or nine times faster than scalar execution. This result confirms the effectiveness of vectorization techniques and applicability of vector processors to Prolog execution and to symbol processing applications.

Rule-Based Building-Block Architectures for Policy-Based Networking

We developed two rule-based building-block architectures, i.e., pipe-connection and label-connection architectures, for describing complex and structured policies, especially network QoS policies. This study focuses on the latter. The relationships or connections between building blocks are specified by the data flow and control flow between them. The data flow is specified by tags, including virtual flow labels (VFLs), which are data attached to “outside packets.” The control flow can be classified and specified by four control structures: concatenation, parallel application, selection, and repetition. We have designed fine-grained and coarse-grained building blocks and methods for specifying data flow and control flow in differentiated services (Diffserv), and implemented the coarse-grained ones in a policy server. Two cases of building-block use are described, and we concluded that there are five advantages of building-block-based policies, i.e., expressibility, uniform semantics, simplicity, flexibility, and management-task-oriented design. We also developed techniques for transforming building-block policies into executable ones, which are called policy division and fusion.