/ Tetsuhiko Ikegami / Vice President
/ Leo Nagamatsu / Associate Professor
/Akira Fujitsu /Assistant Professor
/ Tongjun Huang / Research Associate
ISTC consists of the Information Processing Center and the University Library. ISTC's main purpose is to provide information services through computer facilities and library services.
Refereed Journal Papers
The performance of the classroom computer facilities is evaluated. On class periods when students use computers connected to the network, it is possible to access to one server from many computers simultaneously. For evaluating the peak traffic behavior of the whole system including 100Mbps network environment and file servers, we proposed a method to generate loads by executing processes with finely synchronized on many client computers. From the result of evaluation, we confirmed the effectiveness of this method, and discussed to verify the scalability of current facilities by file transfer rate.
Refereed Proceeding Papers
Leo Nagamatsu., A Bulk Scanning Method for Uncared-for Processes on Classroom UNIX Workstations. Proceedings of 7th International Conference on Parallel and Distributed Systems(ICPADS-2000) Workshops on Industrial Applications on Network Computing (INDAP2000), pp. 409--414, July 2000.
In the University of Aizu, quite many UNIX workstations for student classroom facility are provided. Students -- some of them are very beginners, often logged out a host with unintentionally left processes. And some of these processes may waste processor resources. In high load averaged hosts, such aimless running processes are often found. We developed a tool for periodical search for such processes. It is to obtain basic information for management and for taking an action to keep hosts well. In our university, user files are placed in the central file server and on-demand mounted when a user starts to use a client host. With this configuration, it takes seconds for starting up a user environment. This may be a reason of long response time of remote command invocation. We designed the tool with a pipelined parallel processing, and it can collect process information of each host with high throughput. Achieved average processing time for a host is less than one second, while scanning more than 400 workstations. Also the property of abnormal processes found by this tool is discussed.
With the increased complexity of software and requirement to non-stop feature of services, it's important to establish an effective method for on-line or runtime program updating. While in the domain of hardware, the in-operation replacement of parts -- known as ``hot swap'' -- is common. To establish such ``hot swap'' method for the software system, as well as the way to divide whole system into basic modules / to re-organize them are discussed. On facilities available in traditional UNIX type operating system, such as dynamic linking or process management, prototype routines are prepared. The feasibility of a testbed which employs a version manager included in a simple example application is shown.
With the increased complexity of software and requirement to non-stop feature of services, it's important to establish an effective method for on-line or runtime program updating. In this paper, the possibility of runtime / on-line modification of a working program under some restrictive assumptions are shown. Even with such limitations, the basic idea is effective, we demonstrated it by constructing examples on a simple user application -- replace a sub-function by dynamic linking / unlinking, and process divide and merge procedure while keeping outside interfaces or connected channels.
Unrefereed Papers
On estimating system configuration parameters such as cache size, from program execution time, it is known to use scalability behavior of program executions on various access ranges of the memory. In this paper, we propose a method to reduce cycle of such measuring by dynamic code rewriting and high (nano-second) resolution timing measurement. By this scheme, the measurement speed becomes about 90 times faster than previous one, and reduces the fluctuation of experiment results caused by other system activities.
Academic Activities
Others