I also thought of the difference between distributed systems and parallel processing. In particular, a distributed system is usually on the network and consists of multiple machines. The clocks may not be synchronized (so no common physical clock could be a serious problem--- and techniques like vector clocks are used). But for parallel processing, it has a stronger coupling (e.g. multiple CPUs on one machine, like the ones above), with shared memory. It puts a higher emphasis on the improvement in computation power, so that with multiple cores we can run more than one program at a time now!
axiao
Another difference might be failure cases. If we have a distributed system and one node dies, then a robust distributed system could detect this through something like a timeout and then reschedule the work of the failed node to another node.
In the case of using multiple cores, if a core dies then the processor can't just spin up a new core to use. Furthermore, I'm not sure if the processor can even function properly if one of the cores burns out, so a single core dying could maybe take down the whole processor.
thunder
From my understanding, a distributed system involves machines that actually locate in different geographical locations and link with each other through network. The topic focuses more on the communication mechanism between different machines. The motivation is more about how to enhance the system’s consistency, reliability and availability when machines are talking through network from different locations. However, parallel computing involves multiple processors which can be located one machine or multiple machines. And the goal is more about how to partition a task and run the subtasks on multiple processors in order to achieve better efficiency.
I also thought of the difference between distributed systems and parallel processing. In particular, a distributed system is usually on the network and consists of multiple machines. The clocks may not be synchronized (so no common physical clock could be a serious problem--- and techniques like vector clocks are used). But for parallel processing, it has a stronger coupling (e.g. multiple CPUs on one machine, like the ones above), with shared memory. It puts a higher emphasis on the improvement in computation power, so that with multiple cores we can run more than one program at a time now!
Another difference might be failure cases. If we have a distributed system and one node dies, then a robust distributed system could detect this through something like a timeout and then reschedule the work of the failed node to another node.
In the case of using multiple cores, if a core dies then the processor can't just spin up a new core to use. Furthermore, I'm not sure if the processor can even function properly if one of the cores burns out, so a single core dying could maybe take down the whole processor.
From my understanding, a distributed system involves machines that actually locate in different geographical locations and link with each other through network. The topic focuses more on the communication mechanism between different machines. The motivation is more about how to enhance the system’s consistency, reliability and availability when machines are talking through network from different locations. However, parallel computing involves multiple processors which can be located one machine or multiple machines. And the goal is more about how to partition a task and run the subtasks on multiple processors in order to achieve better efficiency.