What is the difference between distributed arbitration and centralized arbitration?

Question

What will be an ideal response?

Answer 1

In a bus system with centralized arbitration like the VMEbus, one device or module is responsible for
prioritization. Each device that may request the bus sends a message to the arbiter. The arbiter uses an
appropriate mechanism to select the winner if there are several competing devices (e.g., round robin or a simple
priority). The arbiter then invites the winner to take control of the bus.

In distributed or geographic arbitration, there is no central arbiter. Each device capable of requesting the bus
takes part in the arbitration process. Geographic arbitration can be implemented by daisy?chaining. All requesters
are chained together in a line. Each requester has an input and output. Suppose the requesters are chained with
the highest priority device on the left and the least on the right. If the leftmost device requests the bus, it passes
its message to the bus and gets control. If the next device on the right asks for the bus, the message must pass
through the highest priority device – if that device wants the bus, it takes it. If it doesn’t want the bus, it passes
the message on. So when a device wants the bus, its request ripples along the chain from left to right. The
message gets through only if no one else to the left wants the bus (in the VME bus, it is the grant that ripples
along the bus rather than the request – but the principle is the same).

The NuBus uses distributed arbitration by allowing any device to put a code on the arbitration bus. If a lower
priority device requests the bus, it detects that a higher priority device is also requesting the bus and it then
defers by removing its request.