rick@cs.arizona.edu (Rick Schlichting) (06/11/91)
Richard D. Schlichting (rick@cs.arizona.edu)
Associate Professor
Department of Computer Science
The University of Arizona
Tucson, AZ 86721, USA
[This report represents the personal opinion of the author, who was on
sabbatical in Japan from Dec. 1989 through July 1990. The sabbatical
was supported in part by grant INT-8910818 from the NSF U.S.-Japan
Cooperative Science Program.]
---------------------------------------------------------------------------
On Tuesday, I traveled to the west side of Tokyo to visit
NTT Software Laboratories. I arrived at the closest station --
Mitaka on the Chuo line -- at about 9:30am and from there took a
taxi to the labs.
Once I arrived, the guard phoned my host, Dr. Naohisa Takahashi, who
came to sign me in. The lab itself consists of about 10 relatively new
buildings in a pleasant setting. Takahashi led me to his office,
which is in one of the main buildings. His background in mainly in
functional languages, but recently he has been a member of a distributed
computing group and working on replay-style debugging of concurrent,
functional programs.
For about 1/2 hour or so, he introduced the labs to me, overviewing
the structure and groups. He then took me into a lab with (mainly)
Suns to meet two other researchers. One, Ken-Ichiro Murakami, first
described NTT's internal network to me. The system is built on TCP/IP
and is consciously modeled about Stanford's, from which they apparently
received guidance. Among the interesting points made:
-- Compared to the U.S., there is a general lack of network expertise
in Japan.
-- A fair amount of the network equipment they use is American
(e.g., Cisco routers) since theyt have been used more and hence
have been debugged. Good technical support is also an issue.
-- They consider network connectivity to the U.S. as an important
part of general computing infrastructure. They were the first
Japanese affiliate of CSNet, to which they are connected using
two separate commercial transport services.
-- Their efforts at expanding their international computer networking
is hampered somewhat by regulations put forth my the Ministry of
Telecommunications and Post. The basic problem is that they are
not authorized as an international carrier, and so are prohibited
from transferring outside (i.e., non-NTT email) across the Pacific
over their leased lines. They expect some regulatory relief in
the next few years.
Following that presentation, another engineer named Toshiharu Sugawara
demonstrated a heuristic-based Cooperative LAN Diagnostic Expert System.
(A paper on this appeared in the IEEE Phoenix Conf. on Computers and
Communication last year.) The basic idea is that a person having network
troubles (e.g., unable to connect) would use the program to determine the
problem. There is one instance of the program running on each network
segment and the pieces cooperate as needed to determine the problem.
The basic rationale for the development of the program is the lack of
enough trained technical people who can be relied on to manually
diagnose the problem. My understanding is that the program is still
in development and is not yet used on a daily basis.
Next, we went to lunch at the Tsuken Club located in a different
building. The building appeared to house a general cafeteria, but we
ate in a private dining room. Joining us for lunch was Prof. Sam Toueg
from Cornell, who had arrived in Tokyo the previous Friday to start a
month long visit to NTT. Apparently, an NTT researcher had visited
Cornell for a time the previous year, and so they in turn had invited Sam.
It was of course nice to see Sam again, and the entire group of us had
a quite enjoyable conversation over a very excellent continental-style lunch.
My presentation on Psync came after lunch, followed by a presentation
and demo of a software development environment for communication software by
a Dr. Haruhisa Ichikawa. The basic goal of the work is to perform
automatic generation of communication software from specs, especially
in an incremental fashion as the specs evolve over time. The system
includes a graphical display of the specs, which were written in a
standard protocol description language The system also detects problems
with updated specs such as inconsistencies and the like. The system runs
on Sun workstations and is being done jointly with British Telecom.
Although I am not an expert in this area, I found the demo fairly
impressive. In response to a question, however, Dr. Ichikawa seemed
rather skeptical that this would be adopted to develop real-life
communications software at any point n the near term; he essentially said
that development departments have their own established methods that would
be very difficult to change.
The final work described during my visit was a debugger for
parallel and distributed programs. This research was being
directed by my host Dr. Takashi, and the presenter was a young
research engineer who was collaborated on the work named
Yoshifumi Manabe. The system is based on a replay technique and the
use of global predicates to establish breakpoints. Among other things,
they have purused what appears to be some fairly theoretical work on
what type of global predicates are feasible for use in such debuggers.
He gave me a couple of papers (including one in English) that describe
the work.
Overall, the visit to NTT Software Labs was very enjoyable.
My impression from their presentations is that all of the various
projects are technically sound, and that they understand well the
basic issues and background. I also believe that they grasped well
and appreciated the material on Psync that I presented; this includes
the fault-tolerance aspects, which is especially impressive since, as
far as I could tell, they are not doing any work specifically in
that area. I also got the feeling that NTT encourages international
cooperation and publication (their visiting researcher program is
also evidence of this.)
[Addendum added June 6, 1991]
The LAN diagnosis system mentioned above is now being used occasionally
to fix network troubles in the Lab, and has proved to be a useful tool.
Dr. Takahashi's current research interests are focusing on the debugging
of parallel programs with shared data objects.
----------------------------------------
Dr. Naohisa Takahashi
Senior Research Engineer
Software Research Laboratory
NTT Software Laboratories
Nippon Telegraph and Telephone Corp.
9-11 Midori-cho, 3-Chome
Musashino-shi, Tokyo 180 Japan