When the project started in 1991 we only had the Alliant in Dwingeloo to reduce our datacubes. On this machine most of the early reduction software (the N-series reduction package written by Wim Brouw and the Automatic Batch Facility written by Ger van Diepen) was tested.
It was clear from the start that sharing this computer with the rest of the astronomers at Dwingeloo was not feasible. There was not enough diskspace available (2.5 MByte total) and also the processing speed wasn't impressive. A full reduction of 1 mosaic would take about 3 weeks on an empty Alliant (estimates were made when we didn't have a full reduction scheme, so this was a very optimistic guess, 5 to 6 weeks is a more likely estimate). As a temporary solution all other users migrated to the workstation environment that was just being installed. This gave us at least the possibility to use diskspace without the chance that another user had allocated this during reduction.
Because we didn't start mass-production (second half of 1992) diskspace was not a very big problem. Tests in Groningen revealed that an HP730 workstation was 4 to 8 times faster with the N-series programs than the Alliant, depending on the type of program being used (this is a result of differences in CPU power and I/O bandwidth). When Alliant stopped to exist and no new developments (like upgrading the Operating System, expand network capabilities) were expected, the decision to buy an HP730 workstation dedicated for our project was made. The machine could reduce a complete mosaic in 1 week (also an old estimate), which was our aim from the beginning in order to keep up with the datastream from Westerbork and also because we want to complete the program before 1995 when the WSRT will be upgraded with the new digital backend. This will allow us to do follow up work as soon as possible. The new workstation arrived early 1993 complete with 4 GBytes of diskspace, a DAT recorder, a CD-ROM drive and an X-terminal. Porting the software from Concentrix to HP-UX was done without much problems.