The following doesn't really address how MTS came to two sites in Brazil, but it does give a little information about the two sites.
Item 582 16:55 Apr23/80 16 lines 5 responses
BOETTNER, DONALD (UM)
Addresses for Brazilian Installations
According to Jeff Ogden, who just returned from there, the correct
addresses for the Brazilian installations are as follows:
EMB: Stan Angilley
Flora Program Supervisor
CNPq - S.I.P.
W3 Norte Quadra 507 Bloco B
70 000 Brasilia D.F.
RIO: (This one same as in MTS Newsletter 57 page 2 list)
Julio L. Botelho
AV. Wenceslau Bras 71
CEP 22.290 Rio de Janeiro - RJ
Item 587 21:25 Apr29/80 117 lines 5 responses
OGDEN, JEFF (UM)
What I did for my spring vacation - MTS in Brazil
As some people already know from Don's item in MTS:FORUM
I've just returned from a 15 day visit to the MTS installation
in Rio de Janeiro. The trip was made at the invitation of Julio
Botelho and CBPF/CNPq. They paid for my flight down and back as
well as a hotel room and meals during my stay. In return I gave
several informal talks to their staff and one talk to their
users about MTS.
The RIO installation is the second MTS installation in
Brazil. Currently it is a department (GPD) of an institute for
the study of physics (CBPF). CBPF is in turn an institute of a
national government research organization (CNPq). CBPF is
located next to the University of Rio de Janeiro, but it has no
formal ties with the university. However, CBPF and the other
CNPq institutes do accept students who wish to continue their
studies and they do offer graduate degrees. In addition to
providing computing services to CBPF the RIO installation also
provides services to other CNPq institutes, primarily the
National Observatory and the mathematics institute.
The RIO installation is small. They run an IBM 370/145
with 127 pages of real memory, 10 locally attached 3277
displays, two hardwired terminals and two RJE stations connected
to a 3704 emulating a 270x. They have 4 3330-I disks connected
to one control unit on two channels and three 2314 disks
connected to one control unit on one channel. One 3330 is used
for paging and a 2314 is the HASP spool pack. They have no 2301
or 2305 paging devices and one of the two channels used for the
3330s is also used by the 2314s and the magnetic tape equipment.
There are six people on the programming staff including
Julio. Two of the programmers are experienced in assembly
language and the rest are learning. Everyone on the programming
staff and all of the users that I met were very enthusiastic
about MTS. There are some 600 ccids in the accounting files,
but I don't have a good feel for how many individuals this
They are running a hybrid system made up from a special
version of D4.0 for small machines plus selected updates from
D4.1, D4.2 and various RD tapes. In particular they have
installed the CIDA updates from SFU. Because of they way the
IPLREADER and RAMROD work the entire resident system must fit
into real memory. To keep within this limit everything that can
be loaded by PISTLE (and several things that I wasn't sure could
be loaded by PISTLE, like the OPERATOR job) have been moved out
of RAMROD and into SEG2:S2FILES. Even with these changes the
system is within one page of the real memory limit. This means
that the only updates that can be applied to their system are
ones that don't increase the size of the system. Applying
updates using *CDUPDATE is also a bit of a pain on such a small
paging bound system and they are eagerly awaiting the plus
version of *CDUPDATE from UBC. They don't have *ASMH, so most
recent updates have required at least a little modification to
work under *ASMG and some updates have proved almost impossible
to convert. I tried to make it clear that without *ASMH this
problem would only get worse in the future.
The system runs with from 8 to 12 active lines and zero or
one batch jobs using between 900 and 1100 virtual pages during
the day. Maximum paging activity is about 22 page-ins per
second. Processor idle normally runs between 40% and 60% with
the PDP using about half of the CPU time. To say that the
system is badly paging bound is an understatement.
The default global time limit in batch is 1 min. and the
default local time limit for LSS is 10 secs. During the day it
is not uncommon for all of the terminal users to be in LSS. LSS
is used more to encourage batch use than to balance the terminal
and batch load.
During the next year the RIO installation will become the
major part of a new CNPq institute that will provide computing
services to the other CNPq institutes and also do research in
computer science. As part of this change they hope to replace
their 370/145 with a 3031 or a 370/158 with 2Mb of main memory.
With luck this will be done early in 1981.
The original MTS installation in Brazilia still exists and
I talked to Stan Angilley on the phone while I was in RIO. That
machine (also a 145) how has a full megabyte of storage and Stan
is working on installing the D4.2 system. EMBRAPA has
apparently lost some or all of its funding and they are selling
time (VS1 time I'd guess) to make ends meet. Stan didn't say so
directly, but the impression I got from talking to him and Julio
is that very little MTS work is being done in Brazilia other
than the work on D4.2 by Stan. Stan wasn't sure if he would be
able to attend the RPI workshop or not.
Stan and Julio both mentioned that they have been having
problems getting shipments of tapes from Michigan. It seems
that some of the shipments have arrived at ports of entry other
than RIO or Brazilia and that the shipments must be claimed at
that port of entry. This is a little like having something sent
from UBC to UM arrive in Miami and having to send someone from
Ann Arbor down to get it. There doesn't seem to be any trouble
with small tapes sent through the mails and clearly marked as
having no commercial value.
Both Stan and Julio think that they have solved the various
bureaucratic problems that have delayed the payment of their
installations' part of the MTS distribution costs. The payment
from the RIO installation should be on its way now and payment
from Brazilia should be on its way as soon as UM can send a new
bill that doesn't include Stan's name.
Both Julio and Stan have received tentative inquiries about
using MTS from several organizations. If these inquires prove
to be serious they will ask the organizations involved to
contact Michigan directly. I agreed to ask Dr. Finerman to
write Stan and Julio a letter outlining in general terms
Michigan's views on how such inquiries should be handled. I
think that the RIO installation would like to have a second MTS
installation near by if only to provide a bit of moral support
and a larger base of people with MTS experience.
And part of a December 2010 note from Jim Bodwin:
I visited the Rio MTS site to install the latest release under
VM/XA. They had trouble getting disk controllers there so they had
an old 4381 (4341?) that had an internal disk controller that was
normally connected to the CPU via 2-foot channel cables. At Rio
they had the cover off the machine and ran channel cables from
their real CPU (3081?) to the internal controller on the 4341.