In 1972 I was the Development Manager at the IBM Program Information Department, called PID. This department was responsible for distributing software products to U.S. Customers and to sister Program Libraries in Tokyo, Paris, Toronto, and Rio de Janeiro. My boss, Marty Kloomok called me early one morning to tell me the place was burning down. The following is a copy of a brochure we published after we recovered from the fire. To view the photos in full size click on the photo thumbnails and then to return to the article click on your browser back button.
The
IBM Program Information Department fire of September 1972 made headlines.
As an IBM Customer, you quite naturally have far more than a passing
interest in that disaster.
Well,
we're a data processing installation, too, using IBM equipment... and we learned
a great deal from the fire. The
purpose of this folder is to tell you what happened and what we did about it, so
that you can benefit from our experience.
The
PID resource
Although
you are probably familiar with us and our operation, a few
words of background
may be helpful in putting the event into context,
At
the time of the fire, the Program Information Department was located in
Hawthorne, New York.
We
are now fully operational in Mahwah, New Jersey, about 30 miles from
Hawthorne.
Our
mission is to distribute all programs, which are available from IBM.
Shipments are made to all customers in the United States.
This distribution includes supplying master copies to five other IBM
distribution centers around the world. Thus,
PID is a critical resource to IBM and plays a vital role in serving
all IBM customers worldwide.
To
give you an idea of our activity, on an average day we ship 1,500 program orders
and supporting materials to domestic customers.
Each of these orders is accompanied by associated documentation, and is
suitably packaged for mailing. Accordingly,
PID is required to store large amounts of program documentation and packaging
materials.
To
perform our mission, three key elements at PID
constantly interrelate. The first
is the product we distribute-copies of our Program Master Library.
Second, our Customer Master Record files, and In Process Order Data.
And finally, our own operating systems, management programs and
production programs.
On
Sunday, September 10, 1972, at 12:15 AM, a smoke detector went off alerting the
security guard. Within minutes, he
located a fire in a large storage room on the lower level of the building.
The flames were already too extensive to be controlled with hand
extinguishers. The smoke detector
also set off an alarm in the nearby fire department, and firemen with equipment
arrived on the scene at about 12:25 AM. Before
the fire was extinguished, more than 200 firemen had assisted in fighting a
12-hour holocaust hot enough to buckle steel girders, make plastic butter out of
keyboards and tape reels, and reduce tons of paper to ash.
The
storage area, where the fire began, contained packaging materials,
shipping
cartons, cases of 80- and 96-column cards and pallets of program documentation.
Although the packaging materials, cartons, and some cards and manuals
were entirely destroyed by the fire, much of the card stock and tightly bound
manuals were only charred around the edges; they were ruined by the water and
foam used to extinguish the fire.
The
storage area also contained copies of the master tape library, kept in steel
cabinets within a concrete block vault, protected by a co2 extinguishing system. The tapes within the vault were not in any way damaged by the
fire, but the bottom tier was soaked by water seeping into the vault.
The
computer room, which was located directly above the storage area, received its
major damage above the entrance of the air-conditioning ducts from the basement.
At this point, the heat was sufficient to collapse the armor ply floor
panels (one inch plywood core sandwiched between two thin gauged steel sheets). The framing and pedestals also collapsed.
To illustrate the heat intensity under the raised floor, the aluminum
caps on the pedestals melted, which would require a temperature of over 1000º
F.
Although
the computer room had been transformed into a hot oven, the fire damage in this
room was restricted to an area directly exposed to flames from the ducts from
the basement. In the flame-affected
area, two units were consumed by flame, a 1442 and a 2311. A second 2311
adjacent to the first one was ruined by excessive
radiated heat.
A System/3 unit was partially destroyed by fire, which entered the unit
from the cable hole openings. However,
the fire did not spread to other sections of the System/3.
Most of the other data processing equipment in the room was damaged, not
by actual flames, but by heat and smoke.
Although
the room contained a significant number of tapes, some lined up on shelving, in
cabinets, or on low flat carts, a large percentage of them appeared to be
intact. There was much more damage
to the reels themselves. Some of
the reels high up on the shelves and near the floor (on dollies) showed the most
damage.
We
had suffered a great loss; our business was heavily impacted.
First
steps to recovery
The
facilities and the data processing equipment sustained heavy damage,
but the
real story of the PID fire is in its aftermath ... the recovery.
By
3:00 AM Sunday, the morning of the fire, the IBM PID management team was at the
site. By 4:00 AM, a temporary
control center was set up in a nearby motel.
At about noon, the fire was under control, and by 2:00 PM, leased space
at Mahwah had been selected as the site to which PID woul
d move temporarily.
By
late Monday, two System/360 Model 65s, which were being removed from Mahwah,
were reinstalled and operational. All
remaining equipment had arrived and was installed by Thursday afternoon.
During the week, tapes were recovered, machines installed, systems
generated, and documentation and supplies ordered and delivered.
Also during the week, the PID assembly and packaging operations were
reconstructed. All our packaging
material and shelving had been destroyed and had to be replaced; however, we
were able to salvage a conveyor belt system and two large packaging machines.
By Friday afternoon, PID was operational.
On September 18, eight days after the fire, customer orders were being
shipped. And on September 29,
nineteen days after the fire, the backlog was clean and PID was completely
current.
All
this was made possible by two factors-emergency planning and employee
dedication.
As
soon as the alarm went off in PID, IBM Security was notified.
They, in turn, put into action an emergency plan, which started by
notifying all executives with direct control over any resource that might be
important for recovery.
Thus,
when the PID Management team convened at the motel emergency headquarters early
Sunday morning, the necessary IBM people were on-site and ready to help.
The
first critical problem was temporary space.
A representative of IBM's Real Estate and Construction Division arrived
with an inventory of all available space within commuting distance of Hawthorne.
Fortunately, the Mahwah building, which had recently been vacated by
another IBM activity, provided a match to PID's requirements.
Equally fortunate for IBM was the availability of two recently
discontinued Model 65s, which were being crated at the site.
Also
present at the Sunday morning meeting was a member of the Data Processing
Division's Product Scheduling department. His
job was to get the next most critical resource-equipment.
Since space was now available, he could go to work immediately.
The procedures that he followed were exactly the same as those used in
the event of a customer emergency. The
installed inventory files were checked to determine what equipment-down to the
feature level-would have to be replaced, and expediters were named in
headquarters and in each of the involved plants.
In
addition to the System/360s, the installation also required System/3s for
generating 96-column cards and 5440 disk packs, 1130 computers for generating
2315 disk packs, and a Model 20 with a special RPQ for producing paper tape.
By
Monday afternoon, Customer Engineers were reinstalling the Model 65s.
Shipments of the other equipment started arriving on Tuesday and, due to
the emergency support of a Field Engineering task force, the PID installation
was, for the most part, recreated at Mahwah by Friday evening-five days after
the fire.
One
other key activity was started at the motel headquarters on Sunday, the morning
after the fire. The divisional
Purchasing department was asked to review all PID purchase requisitions and contact the appropriate suppliers.
By Monday afternoon, emergency orders had been placed for all the
furniture, equipment, packaging materials and paper forms required by PID
to process and ship customer orders.
Recovery
The
PID disaster plan is only a three-page document. All the managers had participated in either drawing up the
plan or auditing it, so that each was familiar with the problems he would face
after a disaster. The fact is that
the enforced discipline of thinking through the plan prior to an emergency
turned out to be more important than the document itself.
With
all of the peripheral support machinery put in motion, the PID management team
turned its attention to the specifics of the plan for the recapture of the three
elements required to resume normal operations:
The Program Master Library
Customer Master Records and In-Process Order Data
PID'S own programming systems
The
Program Master Library (almost 2,000 tape reels), which was kept in the basement
vault, had one of the best backups of all of the PID resources.
Sub-master (working) copies were kept in the computer library just off
the computer room. Another complete
set is kept in Toronto, and a third complete set in Paris for distribution to
IBM World Trade Corporation's customers.
All
80-column card programs and paper tape programs (kept on tape), and all System/3
card and disk programs and 1130 disk programs (kept on disks) were recovered
on-site. A considerable number of
tape submasters were in process in the computer room and were destroyed by the
heat of the fire. Of these, all but
121 were recovered from the basement vault.
These 121 were in the bottom tier of the vault where the water and foam
had seeped in, and rather than take any chances, the backup libraries in Toronto
and Paris were used. (Interestingly enough, many of the water-soaked tapes were
found to be readable after they had dried out.)
While
the Program Master Library was common to the Canadian and European operations,
the Customer Master Record files and PID'S own programming systems are unique.
Consequently, the PID emergency plan called for extensive backup for
these files and programs. The
active files and programs were kept on 2316 packs, which were online.
A backup set on magnetic tape was stored in a fireproof vault in the tape
library. Another set was taken each
week to an off-site vital records storage area.
Recovery was effected from the tapes stored in the vault.
In the case of complete disaster, recovery could have been effected from
the off-site copies.
Once
the vital records were recovered, the recovery of the work in-process was the
final step. The assumption was made
that all in-process customer orders were destroyed and special programs were
written to reset all files by "backing out" in-process data.
All orders were then reprocessed and all customer-owned tapes and disks
on the premises were replaced.
Although
a large amount of program documentation was salvaged from Hawthorne, there was
100% backup of this documentation in the Publications Distribution Center in
Mechanicsburg, Pennsylvania.
It
can happen to you
In
summary, the major factors in our quick recovery were planning and people:
planning to think through the unthinkable and
establish backup and recovery procedures; and people, whose training, dedication
and commitment are a vital requirement in any emergency.
At
PID we have indeed been tested by fire. We
believe that helping you guard against a similar disaster falls squarely within
our mission of customer service. And the experience of the fire enables us to make some
specific recommendations.
Computers
themselves are not combustible and the supplies that are used to support them
are only slightly so. Our computers
were damaged by heat but they never burst into flame. Cards, tapes, disks and documentation in the machine room did
not ignite. Cards stored right next
to burning material charred on the outside but did not burn.
The recommendation is to store combustibles in an area far removed from
computers and supplies, preferably off-site.
The
effectiveness of fire walls in any facility was dramatically illustrated.
And,
perhaps most important, have a plan for recovery. In an emergency,
you may not be fortunate enough to have the
planners around to implement it, so the document should be stored off-site.
It should contain an inventory of items essential to recovery, a list of
alternate sources of supply, and people to contact at the supplier.
The plan should be simple and flexible.
It is more important that the plan identify alternatives than give
detailed instructions.
A
manual, "The Considerations of Physical Security in a Computer
Environment," can provide you with more information on the subject.
It is available from your IBM representative.
IBM
International
Business Machines Corporation
Data
Processing Division
1133
Westchester Avenue, White Plains, N. Y
10604
(USA
only)
IBM
World Trade Corporation
821
United Nations Plaza, New York, N. Y
10017
(International)