Introduction to Concurrency

Introduction to Concurrency

We all want to write good software. We also know that good software is one that is testable, maintainable, reusable, flexible, and efficient. We are going to focus on the efficiency aspect. When we talk about efficiency, we probably think in terms of speed. For example, let’s say we have a program A that performs a task in 60 seconds. And another program B performs the same task in two seconds. We can say that Program B is more efficient than Program A.

How we can achieve efficiency? One way will be to have a faster computer. Unfortunately, this can be expensive and it’s not scalable in any way as there are limits to process and speed. Another option is to take advantage of the power of the different cores of our processor.

Concurrency means doing several things at the same time. For example, if we have a million tasks to do, instead of doing them sequentially one by one, we can do them simultaneously, thus reducing the duration of the program execution.

One way to visualize this is if you have a restaurant with only one cook, then that person is in charge of cooking everything that each client asks for. It is clear that we want to reduce the waiting time for customers. Then we can hire another cook. In this way, both cooks will simultaneously cook the customer’s food and consequently, customers will have to wait less time to be served.

This concept of having a set of tasks and dividing them into several parts that can be performed simultaneously is called parallelism. Understandably, in our restaurant analogy, we were able to achieve parallelism by adding a new cook.

Well, in programming to achieve parallelism, we can use threads. A thread is a sequence of instructions that can be executed independently of other code. Since they are independent within a process, so we can have several threads. And if our processor allows, then we can run several threads simultaneously. When we are able to execute multiple threads simultaneously, then it is called multi-threading. So, parallelism uses multiple threads to perform multiple tasks simultaneously. Therefore, parallelism uses multithreading and multithreading is a form of concurrency.

However, there are other ways to do concurrency. We just talk about efficiency and we associate it with speed. Efficiency also has to do with resource use. For example, if we have a web server, we want to be able to serve as many Web requests as we can concurrently. For that, we need to release threads when they are not in use. We can do this by using asynchronous programming.

Asynchronous programming allows us to use threads efficiently by using premises and threads are prevented from being unnecessarily blocked.

Suppose you ordered a pizza through your phone to have it delivered to your home. They tell you it will take 30 minutes. What will you do in those 30 minutes? Do you just freeze right there waiting for the pizza, or will you do some other tasks in the house while the pizza arrives? So, obviously, you want to make the most out of your time, so you decide to do a few things while waiting for the pizza.

In our analogy, pizza preparation is an operation that is not going to be satisfied immediately. And you are like a thread. Instead of being stuck waiting for the result of the operation, it is better to do all some other tasks.

Meanwhile, in Web applications, this is useful for being able to scale vertically. That is, we can serve more HTTP requests on our web server and each request is handled by a thread if we avoid the thread blocking, then in general there will be more threads available to satisfy new HTTP requests.

Introduction to Parallel Programming

Parallel Programming helps us to divide a task into different parts and work those parts simultaneously. An example might be that we have a set of credit cards and we want to process them simultaneously. Or if we have a set of images and we want to apply a series of filters to each one, we can do this by taking the advantage of parallelism.

The main benefit of parallelism is saving time. Time is saved by maximizing the use of computer resources. The idea is that if the computer allows the use of multi-threading, we can use these threads when we have a task to solve. Instead of underusing our processor using a single thread, we can use as many threads as we can to speed up the processing of the task.

Parallel Programming is very important for systems that must process a huge amount of data. For example, on Facebook, approximately two hundred and fifty thousand photos are uploaded per minute. As you can imagine, it takes a lot of power to process such a high volume of information. However, the processors are not getting much faster because of the physical limitations. What is being done then mainly is to include more cores in the processors. In this way, we can take advantage of parallelism to accomplish more tasks in less time.

It is not recommended to occupy several threads for one HTTP request. If you have a long task to do, then it is recommended to use background services or some server technology.

In C#, we mainly use two tools to work with parallelism. They are as follows:

  1. The Task Parallel Library (TPL)
  2. Parallel LINQ (PLINQ)

The Task Parallel Library is a library that makes life easier for us. When we see parallelism in our programs that TPL (Task Parallel Library) abstracts the low-level detail of thread handling, allowing us to run programs that run parallel without having to work with these threads manually.

On the other hand, PLINQ or Parallel LINQ is an implementation of LINQ that allows us to work in parallel. For example, in LINQ, we can filter the elements of an array. Then with Parallel LINQ, we can filter the same array in parallel. This allows us to use the cores of our processor to perform the evaluations of the elements of the array simultaneously.

There are two forms of parallelism. They are as follows

  1. Data Parallelism
  2. Task Parallelism

In Data Parallelism, we have a collection of values and we want to use the same operation on each of the elements in the collection. The examples will be to filter the elements of an array in parallel or find the inverse of each matrix in a collection.

Task Parallelism occurs when we have a set of independent tasks that we want to perform in parallel. An example would be if we want to send an email and SMS to a user, we can perform both operations in parallel if they are independent.

Just because we have the concept of parallelism, that doesn’t mean we should use parallelism. We will see later that there are times when it is better to not use parallelism because in certain cases using parallelism is slower than not using it.

Introduction to Asynchronous Programming

Asynchronous Programming allows us to handle the threads of our processes in a more efficient way. The idea is to avoid blocking a thread while waiting for a response, either from an external system such as a Web service or from the computer’s file management system.

The optimal thread management provides us with two very important features i.e. Vertical Scalability and a User Interface that does not freeze. Vertical Scalability refers to an improvement in the processing capability of our application.

There are several ways to achieve scalability. One of them is by using Asynchronous Programming. For example, if we have a web application, it will be able to serve a greater number of HTTP requests at the same time by using asynchronous programming. This is because each HTTP request is handled by a thread, and if we avoid blocking threads, then there will be more threads available to process HTTP requests.

When we talk about a UI that does not freeze, we are referring mainly to desktop and mobile applications with which the user will be able to continue interacting even when there is a process in progress. This is because the interaction with the UI is handled through the UI thread. So, if you allow the UI thread to be blocked by waiting for a long task to be resolved, the user will not be able to interact with the application. Using Asynchronous Programming, we can avoid blocking that UI thread.

To work with asynchronous programming in C# we use async and await. The idea is that we can use async to mark a method as asynchronous and with await, we can wait for an asynchronous operation in such a way that the original thread is not blocked.

The normal thing is that the method marked with async returns a Task or Task<T>. The idea of a Task is that it represents an asynchronous operation. In the case of Task<T>, it is like a promise that in the future this method will return a value of the data type T.

Asynchronous programming can be used in any environment like Desktop, Mobile, and Web. Normally we use asynchronous programming when we are going to communicate with external systems. For example, if from our application we have to communicate with a web service, we will want to use asynchronous programming.

This is an I/O bound operation. I/O bound operations are characterized by the fact that their performance depends on communication between systems. This is why asynchronous programming does not improve the speed of the processes since there is no way that from our system, we can make the processing speed of an external system faster. The most we can do is to be efficient in managing our threads so as not to waste resources waiting for IO operations.

CPU vs I/O Bound Operations:

We are already discussed what Asynchronous and Parallel Programming are. It is also important to understand what type of operations both are intended to try to improve.

In the case of Asynchronous Programming, we discussed that it has the specialty to handle the IO-bound operations where IO-bound operations are characterized by communication with external systems. Some examples of IO-bound operations are calls to a Web Service, interaction with a Database, interaction with a file system, etc. Therefore, when we need to perform such kinds of operations, we can consider the use of asynchronous programming to increase the level of scalability of our systems.

When we make a call to an external entity, we have to wait for a response and while waiting for the response, it is productive to free the thread that started the operation so that it can proceed to perform other tasks.

On the other hand, CPU-bound operations are those that are performed primarily using processor power. Here, there are usually no dependencies on external systems, everything depends on our system. If we have multiple CPU operations that are independent, we may want to use parallel programming to decrease the time it takes to perform these operations. Some examples of CPU operations are finding the inverse of a matrix, sorting the elements of an array, etc.

It is also important to understand the difference between IO and CPU bound operations to see what you can consider using Parallel or Asynchronous Programming.

If your operation requires communication with some external system to your program, then it is IO bound and therefore you can consider asynchronous programming. On the other hand, if the operation is done entirely within your program and its execution time depends on the processor, then it is a CPU-bound operation and therefore you can consider using parallel programming.

Sequential Programming, Concurrency, Multithreading, Parallelism, Multi-Tasking:

In the context of concurrency, certain relevant terms are handled. Some of these terms are very similar and the differences between them are often certain. Even if they are used interchangeably in informal contexts, they are not exactly the same. We will look at the concepts of Sequential Programming, Concurrency, Multithreading, Parallelism, and multitasking. Let’s start with the non-concurrent programming model.

Sequential programming: Sequential programming is the one in which the instructions are done one at a time. That is where there is no concurrency of any kind. One of the advantages of this programming model is that it is relatively easy to understand since it consists of following a series of steps in an orderly manner. The problem with this programming model is that sometimes it can be slow.

Concurrency: Concurrency means doing several things at the same time. This is the opposite of sequential programming. The term concurrency encompasses everything related to in one way or another doing several things at the same time. There are different forms of concurrency. We have seen a fundamental concept of threads. We remember that a thread is a sequence of instructions that can be executed independently of our code.

Multithreading: Multithreading is the ability to use multiple threads. It is important to clarify that multithreaded does not imply parallelism, since we can have a computer with a processor that is not multicore and I still can use multithreading. This is because an operating system can provide several threads and execute them sequentially without using parallelism

Parallelism: It is running several threads simultaneously. This requires a multicore processor. Since parallelism uses multiple threads, so parallelism uses multithreading. However, as we said, we can have multithreading without having parallelism. In this case, typically what we have is called multitasking.

Multitasking: With multitasking, we can have several tasks running in such a way that we execute their different threads sequentially, typically with some type of Task Execution System. This is handled at the operating system level. For example, if we have a program A with threads one and two and a program B with threads three and four, and we try to execute both programs at the same time, it could be that the system executes the threads in the order one three two and four.

Introduction to Parallel and Asynchronous Programming

So, it looks like there was parallelism, but there really wasn’t as the threads did not run simultaneously, but in sequence. The computer is so fast that the human eyes could not see that the task was executed in sequence.

Determinism vs Non-Determinism

There are methods where we can predict its result from its input values. If we have a method that takes two integers as input values and returns the sum of the two numbers, then it is clear that we can predict the output value from the input values. If we send 2 and 3, the result will be 5. i.e. 2 plus 5 is seven. This characteristic of being able to predict the result of a method based on its input values we call determinism.

What happens in the opposite case? That is when we have a method where we cannot predict the result. Well, we say then that we are facing a non-deterministic method. A simple example of non-determinism will be the Random class. With this class, we can generate pseudo-random numbers.

Therefore, the output value of the Random method cannot be determined from the input values supplied to its methods. Therefore, the output value of the Random class methods cannot be determined from the input values supplied to these methods.

However, not only with the random class, we have nondeterminism, Parallelism can also cause some sort of non-determinism. Suppose you have a method that processes credit cards and as it processes them writes a message to the console window. If we use sequential programming, we can always predict the order of the messages on the console window. With parallel programming, this is virtually impossible to predict. We know that all operations are going to be executed, but we have no way of knowing the order of execution of the threads that will be in charge of processing the different credit cards. Even if we know that all credit cards will be processed, we cannot predict the order of processing.

Therefore, we must be kept in mind that when we use code in parallel, we will not be able to predict the order of operations until we perform. If you need to have a specific order in the tasks that you have to do, then maybe parallelism is not a good option in your case.

  1. We saw that concurrency refers to, in one way or another, doing several things at the same time. That concept of concurrency encompasses parallel programming and asynchronous programming.
  2. Parallel programming refers to the use of multiple threads simultaneously to solve a set of tasks. For this, we need processors with adequate abilities to perform several tasks at the same time. In general, we use parallel programming to gain speed.
  3. Asynchronous programming refers to the efficient use of threads where we do not block a thread unnecessarily. But while we wait for the result of an operation, the thread gets to perform other tasks in the meantime. This increases vertical scalability and allows us to prevent the user interface from freezing during long tasks.
  4. CPU-bound operations are those that depend entirely on the speed of our processors.
  5. IO-bound operations are those that depend on communication with entities external to our application.
  6. Deterministic refers to the fact that we can’t predict the result of something based on the initial conditions. For example, we can predict the result of a method from its input values. With parallel programming, we will not always be able to predict 100 percent the result of something, especially when we refer to the order of operations of a set of tasks, since we do not control the order of execution of the different threads of the application.

In the next article, I am going to discuss how to implement Asynchronous Programming using Async and Await Operators in C# with Examples. Here, in this article, I am trying to explain the basic concepts of Parallel and Asynchronous Programming.

1 thought on “Introduction to Concurrency”

  1. Guys,
    Please give your valuable feedback. And also, give your suggestions about this Parallel and Asynchronous Programming concept. If you have any better examples, you can also put them in the comment section. If you have any key points related to Parallel and Asynchronous Programming, you can also share the same.

Leave a Reply

Your email address will not be published.