Two models for concurrent programming

Concurrency implies that multiple tasks are happening in parallel to achieve a common goal. Just like communication with a group, we need to communicate and coordinate with the concurrently executing entities.

For example, let us say that we present the previously mentioned word frequency counter via a UI functionality. A user uploads a huge file and clicks the start button, resulting in a long-running MapReduce job. We need to distribute the work among the slaves. To send the workload, we need a way to communicate with them. The following diagram shows the various streams of communications that are required in this system:

If the user changes their mind and aborts the job, we need to communicate the stop message to each concurrent entity, as any further processing is futile.

There are two prominent models for concurrent communications: message passing and shared memory. The preceding diagram shows a message passing model.   

We will first discuss the message passing model, using the celebrated UNIX shell pipeline as an example. Next, we will see the shared memory approach in depth and the problems that are associated with it.