Cray Y-MP EL restoration during 2011
Post this page to popular social media
Each TNMOC project has either a working group or project team assigned to do the work. Working groups are either managed in association with the CCS (Computer Conservation Society) or solely within the Museum.
Below you can follow the restoration of our Cray Y-MP EL Super-computer by Delwyn Holroyd and Adam Bradley.
12/06/2011 Update from Delwyn Holroyd
I re-installed the Memory power supply in the machine and found that it does indeed now work, although Phil didn't find any fault with it. Clearly there is something intermittent which may return one day.
Having confirmed the total current draw of the machine would be less than 13A (just) I powered up all the modules one by one. After a couple of false starts the system booted into Unicos: Cray's version of Unix which is the operating system on all but the very early Cray machines.
There are a few configuration problems to sort out, and ideally we want to re-organise the disk storage so that not all 8 drives are required, but the system is now basically running. I tested the network interface and was able to ping another machine each way.
Upon logging in I was presented with a somewhat out of date message warning of intermittent service on the Cray on Wednesday 9th July due to planned upgrade work - we guess this must refer to 2003!
04/06/2011 Update from Delwyn Holroyd
Phil H has looked at the Memory PSU and found it to be working again! Internal examination showed that the supply can be run from 230V AC, which will make investigation a lot easier should it play up again when re-installed in the machine, which is the next step.
07/05/2011 Update from Delwyn Holroyd
IOS (Input/Output System), Memory and CPU power supplies bench tested using one of the bulk converter rectifiers, Johan's dummy load and a collection of other power resistors. Machine then reassembled and disk tray power supplies tested in situ on dummy load. No problems found. However the Memory power supply died when turned on in the machine on full load. Phil H is currently looking at it.
Booted the IOS successfully!
Indications are we might just be able to run the machine on a 13A plug as the projected full load is ~11A but this needs to be verified once the Memory power supply is back in action.
12/03/2011 Update from Delwyn Holroyd
Work is now well under-way. All disks have been read and secured. Power supply testing is on-going. The 12V supplies have been tested on their full load (fan trays) and the 380V DC bulk converters on a light dummy load.
5/03/2011 Update from Delwyn Holroyd
We had the machine powered on for most of the day to gain confidence in the 12V supplies, fans and line filtering components of the power system. Meanwhile the two 380V DC rectifiers were tested out of the machine on a dummy load (see picture on right). Running at about 50W output this is a fraction of what they would normally deliver in the system, but sufficient for a confidence test. We can now begin to test the other power supplies in the system one at a time.
26/02/2011 Update from Delwyn Holroyd
We re-assembled the machine with the 12V supplies but without the 380V DC supplies. It was now time to power on to give the 12V supplies, fans and control panel a proper test. It certainly makes a lot of noise! The photograph shows the control panel lit up, and the big red emergency stop button which was also tested.
19/02/2011 Update from Delwyn Holroyd
We started inspection of the power supplies and wiring in the machine. The power system is unusual by comparison with a modern server: incoming AC mains is first rectified to 380V DC and then fed to individual supplies for the processor, memory, IOS (Input/Output Subsystem) and two disk trays. There is also a "power ride-through" module or capacitor box (see picture on right). This acts as a UPS on the high voltage DC rail, capable of sustaining the machine for up to 10s of AC power loss. The charge stored in the capacitors would be lethal and so a couple of electric heater elements are provided to discharge the capacitors when the red 'emergency stop' button on top of the machine is pressed.
There are also two 12V supplies used to run the fans and the control panel. We removed these supplies and tested them both into dummy loads.
12/02/2011 Update from Delwyn Holroyd
Over the last two weeks all of the disk storage has been copied. This consists of 8x 3GB full height 5.25" SCSI hard disks and a 200MB system disk. We used a PC running Linux fitted with suitable SCSI adapters to do the copying (see below). The Linux 'dd' command was used to take a complete dump of each disk.
05/02/2011 Update from Delwyn Holroyd
Work has now resumed on recommissioning the museum's Cray Y-MP EL, an Entry Level Cray system dating from the mid 1990s. One of the first things we discovered about the machine is that it has actually been upgraded to an 8 processor EL98 model, with 2G of memory. The machine also contains 24G of disk storage. This makes it easily the most powerful of the museum's exhibits.
The machine has not been powered up for some time, and is thought to have suffered from some power supply problems, although it was believed to be working before delivery to the museum. It is in generally good physical condition, so the work to be done is mainly inspection and testing.
There was some physical damage to the flap on the control panel, but this was repaired last year.
The photograph on the right shows the business end of the machine - at the bottom are the processor and memory cards, and the smaller rack at the top holds the I/O processor, for connection to network, console and disk storage. Right at the top are located two tape drives and the system disk.