Theoretically, a protocol can define where a machine has some remote calls on another machine (or more than one) , And where in any part of the process, if any of the machines (or operations) fail, or the communication drops, everything has rolled back? (As soon as the database can be done)
I say at the hardware level, because always it says that no nuclear processor operation (without testing and setting) is a nuclear operation (a very important component of the transaction) Can create)
But from now on we are talking about many machines, it is not flying.
For example how difficult it would be to say: I have a protocol to issue an order on a remote machine, and a reaction is returned. It is possible that the method is called, but during the transit of the reaction, the connection dies. It can also be very well that the machine that performs the operation, thinks that everything was fine, but the receiving machine has never received the answer.
Adding AEC does not help, transit
Interested in reading the thoughts of others (and to know that 27 years ago some professors had already come up with a solid concrete solution).
R
Yes, this problem already exists (more or less) Has been done :)
What you are searching for is that you are searching.
In transaction processing, database and computer networking, the two-step committed protocol (2 pcs) is a type of nuclear commitment protocol. It is a distributed algorithm that coordinates all processes that do not participate in distributed nuclear transactions, transactions (roll back) system failures (either process, network node, communication, etc. failure) In many cases of incorporation, the protocol achieves its goal, and thus it is widely used.
No comments:
Post a Comment