Department of Engineering

IT Services

The Central System machines

Computing Models

Groups of computers can be organised in various ways. Options include

  • Mainframe - in this kind of set-up a powerful central machine has many terminals connected to it. All the programs and user's files are on the central machine. Years ago this was a popular option. It's still used with supercomputers
  • Individual Machines sharing resources - the computers are stand-alone machines each with their own disc of user-files. They might share a few resources - printers, etc.
  • Networks - a distributed system, with much use made of client-server processes. That is, the machines ask each other for services (computational power from a compute-server, files from a file-server, etc).
  • Cloud computing

At CUED all these models are used, often in combination. The Central system mostly follows the network model.

You are not alone...

tw104tw100needle Suppose you sit at the tw104 terminal in the DPO and log in. What happens? Let's first consider tw104 itself. It's a PC running linux - a version of Unix. Your files aren't on tw104 though - they're on a machine called file-serv, which acts like giant hard disk connected to all the terminals. So every time you access one of your files it has to be transferred on the internet from file-serv. For security reasons you can't log into file-serv - besides, it lacks most of the programs you usually use. If you edit a file, the editor will get your file from file-serv.
tw104 is fairly fast but if you want to use more powerful machines (they're collectively known as ts-access) you can access it from tw104

But why?

Using 3 machines and the internet just to run programs might seem an odd and wasteful way to operate but it has several advantages

  • Efficiency - Each machine can concentrate on doing its specialised job well. The tw104 terminal can be optimised for graphics; the ts-access compute-server can be optimised to run programs; and file-serv (a file-server) can focus in rapid file-handling.
  • Cost - we can arrange things so that the machines we need the most of can be as simple and cheap as possible.
  • Security - the expensive machines and those that hold valuable data can be kept in a safe place
  • Flexibility - when you login at tw104 you by default use tw104's processor to run programs, but you can also - simultaneously - run programs on other machines too. If in a tw104 terminal window you type
        slogin -X ts-access
    that gets you into the ts-access machine. If you then type
        xclock
    the xclock program will run on ts-access and send its graphics to the X program running on tw104 for display.

The structure of the system is roughly this -

[Teaching System]

Each compute-server/linux-server is a little like a mainframe (or the cloud). Here we call the compute-servers Linux Servers and group of machines like tw101-tw114 a cluster.

gate is the collective name for a group of machines that are just like ts-access except that they have no terminals connected to them by default - you have to use slogin or something similar to remotely login to it. You can do this from a DPO terminal or from distant machines, because gate (unlike ts-access) is visible from outside CUED. They lack many of the programs that the ts-access machines have.

You can't login to the infrastructure machines. They provide services (backups, mail-filtering, etc).

Even more machines

tw104 uses other CUED machines too

  • it synchronises its clock with ntp0.eng.cam.ac.uk or a similar machine)
  • it gets permission to run matlab from matlab-lmserv.eng.cam.ac.uk (similarly for some other programs)
  • to print it uses print-serv.eng.cam.ac.uk

These machine names are aliases - the actual hardware that these services run on might vary from week to week (and many of the aliases could refer to a single machine), but the aliases should always work. For more details look at

Further afield

The machines mentioned so far are within CUED. If you use e-mail however, you're likely to bring the hermes machine(s) into play. hermes is a central university facility - hermes problems are usually beyond our control!