RFC 374 UNKNOWN NETWORK OPERATIONS · INFRASTRUCTURE

IMP System Announcement

An announcement of the new IMP software release and network power failures.

On rfc-editor.org ↗
PUBLISHED
July 1972
AUTHORS
A.M. McKenzie
STREAM
Legacy
RELEASE DATE
July 18, 1972
UPTIME ON RELEASE
30+ hours
DOI
10.17487/RFC0374

RFC text

Original text · RFC Editor View on rfc-editor.org ↗
Network Working Group                           Alex McKenzie
RFC #374                                        BBN
NIC 11099                                       19 July 1972


There are two significant developments in the operation of the network
which should be of interest to all sites:

NEW IMP SYSTEM IS RELEASED !
   On Tuesday July 18, the new IMP software described in previous RFC's
   (#331,#343,#359) was apparently successfully released.  We ran from
   10:00am on that day until Wednesday afternoon with very few network
   problems attributable to the new system.  We are quite pleased with
   this performance, and expect to leave the new system up permanently!!

   We are, of course, aware that in any system of this size, bugs may
   crop up at any time, and we solicit the help of the network community
   in finding them.  We also believe that the new system may have a
   different "feel" to users, both terminal users and Hosts.  We urge
   everyone to have a bit of patience while getting used to any change in
   the "feel" of the network.  On the other hand, PLEASE REPORT ANY
   SUSPECTED NETWORK MALFUNCTION TO THE NCC AS SOON AS POSSIBLE.
   Evidence of bugs, guesses as to what went wrong, etc., are always
   useful in hunting bugs, but don't let the lack of positive proof delay
   the reporting of suspected system malfunction.  The NCC telephone
   number is (617) 661-0100.  Feel free to call collect!

   The one known shortcoming of the new software is the exacerbation of a
   long-standing TIP software problem.  The TIP now often appears to
   neglect sending an Allocate control message for some active
   connection, with the effect that the Server-to-User path becomes
   blocked.  We are giving the solution of this problem the highest
   priority!  (Our current data seems to indicate that this problem is
   most likely to occur when there are several users and when the Server
   Host is a TENEX system.  If you know of other situations where this
   problem has occurred, please let us know!)

   Now that the new system is up, all sites with Model 516 IMPs will be
   contacted by telephone and asked to reset the switches on their IMPs
   as follows:

             HALT INH         - on
             W.D.T.           - on
             AUTO RSTRT       - on
             MEMORY PRTCT     - on

SITE POWER PROBLEMS
   The network has recently been experiencing serious difficulties
   related to the annual problem of more demand for electricity than is
   available during the summer months.  For example:

       During the month of June, between 1/4 and 1/2 of the
       IMP down time was due to site power failure.

       During the weekend of July 15-16, two sites were down
       all weekend due to air-conditioning failure which was
       apparently related to power failures or brownouts.
       This left several sites connected to the rest of the
       net through only one path which, according to Murphy's
       law, eventually failed - leaving nine sites partitioned
       from the remainder.

       On July 18, an underground power transformer supplying
       power to the entire computer room at site 27 failed,
       putting IMP 27 out of action for several hours (anddit
       could easily have been several weeks.)

   Obviously in these cases there is little that either the NCC or the
   site can do to restore service.  We expect these situations to
   continue or even get worse as the summer continues.  Please try to
   restrain your annoyance when these situations arise; we don't like
   them either!




       [ This RFC was put into machine readable form for entry ]
       [ into the online RFC archives by BBN Corp. under the   ]
       [ direction of Alex McKenzie.                      1/97 ]

AI ELI5

On this page
PROBLEMThe ARPANET IMP software was being replaced with a new version, and power failures were disrupting network connectivity.
MECHANISMBBN released new IMP software on July 18, 1972 and asked the network community to report any issues. Regarding power, the memo documented recent failures and encouraged patience.
STATUS UPDATEThe new IMP system ran successfully with few problems, though a TIP allocation issue and summer power constraints remained.

§1New IMP Software Released

On Tuesday, July 18, 1972, the ARPANET released new IMP software described in earlier RFCs (331, 343, 359). The system ran from 10 AM that day through Wednesday afternoon with very few network problems. BBN was pleased with this performance and intended to keep the new system up permanently.

After years of work, the new traffic-switching software finally went live and stayed up. The team was happy.

§2Known Issues and Support

BBN acknowledged that bugs might appear in a system of that size and asked the network community to report them. They also noted the new system might feel different to both terminal users and host computers. The main known issue was the TIP software problem where allocation control messages were sometimes missing, blocking the server-to-user path. This problem seemed most likely when multiple users were connected and the server host ran TENEX.

New software always has bugs. The team asked people to tell them when something feels wrong. One problem: sometimes the new software lost track of resource deals, blocking connections.

§3Site Power Problems

During June 1972, between one-quarter and one-half of IMP downtime was due to site power failures. On July 15-16, two sites went down all weekend from air conditioning failure, leaving nine sites disconnected from the rest of the network through only one path, which then failed. On July 18, an underground power transformer at site 27 failed, taking IMP 27 offline for several hours.

Summer heat was overloading power supplies. Some computers had to shut down, and two went dark for a whole weekend. The team asked everyone to be patient because they could not fix weather.

§4References