I think the first thing to think about is Travellers flying around the galaxy in 100-ton commercial ships for three millennia. There's no way that is possible without small ship logistics being a long solved problem.
As for those solutions, modern spaceflight provides a pretty good model I think. The main computer system on the ISS is built around networks of modular devices. The data bus they sit on is redundant and many of the devices are multiply redundant.
The main computing units are called a Multiplexer/DeMultiplexer (MDM) which is composed of several circuit boards plugged into a backplane. The MDM chassis has power, the I/O controller (32MHz 386SXs with 8MB of RAM ), and other bits. Different functionality is provided by different circuit cards which can be replaced in orbit by ISS astronauts. MDMs process data from sensors and control things like the motors to rotate the solar panel arrays.
Each MDM's controller runs a real-time OS and loads a program/firmware for whatever task it is doing. Because the MDMs are standardized and the expansion cards are replaceable, an astronaut can pull an MDM and replace all the cards with spares and stick it back in place. In a pinch an MDM can be reconfigured so it can replace a unit attached to a different subsystem.
Applied to a Type-S, major ship subsystems would be attached to MDMs all networked to control MDMs. Those can run scheduled/batched jobs and pull in all the sensor data and write it to mass storage. These subsystem controllers and their spares are just part of the mass for each subsystem. The ship's computer would be a more traditional computer running whatever, it might be able to access all the MDMs but it doesn't necessarily have to run them as each have their own control programs.
Again, this setup being in existence for millennia means issues of software compatibility is a long solved problem. Processors of the far future may just natively support dozens of instruction architectures or software is written to bytecode (a la Java) so every processor vendor just includes a low level interpreter(s). Processors of the far future might be more like FPGAs and automagically reconfigure themselves to run whatever architecture the program is compiled for.
As for crew watches, most systems on a ship will be automated with the crew person on watch being the executive authority if things go wrong. While most systems will self-diagnose it would take a human to replace a part. In a full emergency the rest of the crew would be woken up to handle the problem. The one on watch is basically overseeing the automated systems.
As for those solutions, modern spaceflight provides a pretty good model I think. The main computer system on the ISS is built around networks of modular devices. The data bus they sit on is redundant and many of the devices are multiply redundant.
The main computing units are called a Multiplexer/DeMultiplexer (MDM) which is composed of several circuit boards plugged into a backplane. The MDM chassis has power, the I/O controller (32MHz 386SXs with 8MB of RAM ), and other bits. Different functionality is provided by different circuit cards which can be replaced in orbit by ISS astronauts. MDMs process data from sensors and control things like the motors to rotate the solar panel arrays.
Each MDM's controller runs a real-time OS and loads a program/firmware for whatever task it is doing. Because the MDMs are standardized and the expansion cards are replaceable, an astronaut can pull an MDM and replace all the cards with spares and stick it back in place. In a pinch an MDM can be reconfigured so it can replace a unit attached to a different subsystem.
Applied to a Type-S, major ship subsystems would be attached to MDMs all networked to control MDMs. Those can run scheduled/batched jobs and pull in all the sensor data and write it to mass storage. These subsystem controllers and their spares are just part of the mass for each subsystem. The ship's computer would be a more traditional computer running whatever, it might be able to access all the MDMs but it doesn't necessarily have to run them as each have their own control programs.
Again, this setup being in existence for millennia means issues of software compatibility is a long solved problem. Processors of the far future may just natively support dozens of instruction architectures or software is written to bytecode (a la Java) so every processor vendor just includes a low level interpreter(s). Processors of the far future might be more like FPGAs and automagically reconfigure themselves to run whatever architecture the program is compiled for.
As for crew watches, most systems on a ship will be automated with the crew person on watch being the executive authority if things go wrong. While most systems will self-diagnose it would take a human to replace a part. In a full emergency the rest of the crew would be woken up to handle the problem. The one on watch is basically overseeing the automated systems.