/ Hidesada Kanda / Professor
/ Akira Fujitsu / Assistant Professor
/ Tongjun Huang / Research Associate
ISTC consists of both the Information Processing Center and the University Library. A brief outline of these two parts in 1994 is given below.
Information Processing Center
In order to provide a convenient environment for faculty members, staff, and students, ISTC began to provide new support to the AINS (Univ. of Aizu Information Network System) in 1994 as follows.
Up to end of 1994, the AINS had grown into a campus network including about 800 workstations, 2 super computers, 73 network printers, and 192 local printers (Table-1).
University Library
The University Library has been designed to provide both centralized and decentralized services. In accordance with the completion of construction of the University Library and the Research Quadrangles, full-scale service for Library users and management of the research libraries has begun. The statistics on the increase by Fiscal Year are shown in Table-2.
MUSE System (Multimedia University System Environment)
MUSE is a university information system and mainly consists of administration and library systems. Some new functions as shown in Table-3 were added to MUSE in 1994.
Organization
(1) In June 1994, a new member, assistant professor Akira Fujitsu joined our center. (2) In order to solve problems and respond to questions more quickly, the SSB (System Support Base) was opened in the Research Quadrangles.
Refereed Journal Papers
A. Fujitsu (KEK, Tsukuba). Ope. math: Operator product expansions in free field realizations of conformal field theory. Comput. Phys. Commun., 79:278--99, 1994.
We present version 1.0 of the package
"ope.math" for
Tongjun HUANG, Zixue CHENG, Minetada Osano, and Norio SHIRATORI. A decentralized social algorithm for committee coordination problem. In The second international Symposium on Autonomous Decentralized Systems ISADS 95, Phoenix, Arizona, USA, April 1995. IEEE, IPS, SICE, IFAC.
In this paper, we consider an extended committee coordination problem in an autonomous decentralized social environment. The basic committee coordination problem and a distributed algorithm, as a solution to this problem in distributed environment, are presented by K.M.Chandy and J.Misra in. The algorithm guarantees the synchronous and exclusion properties of the problem and is fair (i.e. starvation-free) and makes progress (i.e. deadlock-free). In an autonomous decentralized social environment, besides the above mentioned properties, some social properties, such as individual preference, privacy protection, and stable assignment are required to be considered and guaranteed. In this paper, we extend the committee coordination problem by introducing some social properties, such as individual preference and stable assignment and give a decentralized social algorithm as a solution to this extended problem. The algorithm guarantees not only the synchronous and exclusion properties but also individual preference and stable assignment properties.
LOTOS is a high level specification language which incorporates the Multi-Rendezvous. Multi-Rendezvous is a powerful communication mechanism that allows a set of processes to execute an event in synchronous way. Besides the multi-synchronous property, Multi-Rendezvous has nondeterministic and exclusion properties. It is not a trivial problem to implement Multi-Rendezvous in a network system. The implementation has to maintain the properties mentioned above, be fair and make progress. In this paper, we propose a new efficient distributed algorithm, based on the concept of coterie, for implementing Multi-Rendezvous with $O(N\sqrt{N})$ message passes. Our algorithm is fully distributed and does not use manager processes and auxiliary resources.
Distributed resource allocation is an important problem in developing distributed systems. Many works have been done on the problem. However recent new applications, such as supporting collaborative group work, distributed cooperative development of software, cause some new problems. One of them is that some resources are competed for by several groups of processes. There may exist deadlock or starvation among groups. We call them ``group deadlock'' and ``group starvation''. In this paper, we first define the new resource allocation problem: Distributed Resources Allocation among Process Groups. This problem is an extension of the Distributed Resource Allocation problem. Then we present a new distributed algorithm as a solution to the problem. Our algorithm can allocate resources to groups of processes without group deadlock and group starvation.
In this paper, we propose a new efficient distributed algorithm for implementing Multi-Rendezvous with $2Nm$ message passes for a Multi-Rendezvous, without using auxiliary resources. Here, $N$ is the number of processes which participate in a Multi-Rendezvous, $m$ is the number of possible Multi-Rendezvous within a system. A formal specification of the algorithm in LOTOS is included in the paper.