Computer history: Computer lines at TNO Waalsdorp (1990+)

 

Future computer lines at TNO Waalsdorp

 
At the beginning of 1990, the final report by the IT Policy Committee (ITBC) working group on the replacement of the CYBER 840A was presented to the management. Based on developments in large-scale computing and the need for IT systems at TNO-FEL and TNO-PML, the working group came up with the following recommendations:

  • Phasing out the NOS/VE line by ending the CYBER 840A services on 1 June 1990 and by phasing out the project obligations for NOS/VE services on a CYBER 930.
  • Reduce the number of supported computer lines.
  • Provide a decentralised back-end/front-end concept in addition to the central facilities and services.

The conclusions from this report led to the decision to decommission the CYBER 840A on May 1, 1990. From June to December 1990, the CYBER 930-11 was temporarily upgraded to a CYBER 930-31 to enable a phasing out period. The disk capacity of the 930 was increased to 1.7 Gbytes. The CYBER 930-11 was then put out of use at the end of 1993. In the end, the system provided a competitive trade-in discount.
A technical working group then examined various options for a central Unix server facility, which had to provide the backup and other (network) services for the Convex and the decentral Unix workstations. We investigated systems from HP, SUN, Silicon Graphics, DEC and CDC. It was decided to purchase a Control Data 4380 (MIPS 3840 system) under Unix System V with a jukebox backup system.

CDC 4380 block diagram
CDC 4380 block diagram

In July 1990, a CDC 4380 Unix Server was installed. Soon, the disk capacity was expanded to over ten Gbytes. The Unix Server was used for:

  • pre- and postprocessing system for the CONVEX,
  • processing all Pascal programs, and
  • integer/logic-based computational work.

Moreover, the Unix Server had to provide automatic backup service for all Unix workstations of the laboratory. The Automatic Workstation Backup Service (AWBUS) package, developed by Control Data for Chrysler, was used to provide these services. Initially, we encountered many problems with this facility. First of all, the TNO-FEL definition of a workstation (servers with 1-2 Gbyte disk capacity and up to 10.000 disk files) was quite different from the original components design environment at Chrysler (a couple of hundred Mbyte and a couple of hundreds of files). That caused hardware problems in the automatic Exabyte carrousel robot. After a couple of months of discussions, Control Data replaced the AWBUS system and upgraded the software. As the backup device, an HP 90 Gbyte optical rewritable compact disk robot was installed.

 
The central ICT facilities such as all systems in the central computer room, terminal servers, bridges and the like needed continuous monitoring. Failures and problems with the performance and load of systems, print work, magnetic tape applications, and network communication had to be detected and solved as early as possible. Until 1992, an operator was sitting behind the operator console in the computer room to perform these tasks. The operator could also provide some services to users such as mounting a magnetic tape, tearing off print and plot output, granting user access to systems (account management), and so on. In April 1992, a self-developed automatic alarm system fully took over the monitoring tasks. The system was based on a PC that carried out checks via the network on all central systems and network components. It also checked whether the network connections with TNO institutes PML and IZF still worked. If an action was desired, the system could dial a telephone number with which a beeper went off or a pager call was placed. The warning system was nicknamed XINlX (‘I see nothing’) after Asterix and Obelix. Lights could be dimmed in the computer room.