Blockchain Quick Reference
上QQ阅读APP看书,第一时间看更新

Distributed systems

A distributed system is one in which the application and its architecture is distributed over a large number of machines and preferably physical locations. More simply, a distributed system is one where the goal of the system is spread out across multiple sub-systems in different locations. This means that multiple computers in multiple locations must coordinate to achieve the goals of the overall system or application. This is different than monolithic applications, where everything is bundled together.

Let's take the example of a simple web application. A basic web application would run with processing, storage, and everything else running on a single web server. The code tends to run as a monolith—everything bundled together. When a user connects to the web application, it accepts the HTTP request, uses code to process the request, accesses a database, and then returns a result.

The advantage is that this is very easy to define and design. The disadvantage is that such a system can only scale so much. To add more users, you have to add processing power. As the load increases, the system owner cannot just add additional machines because the code is not designed to run on multiple machines at once. Instead, the owner must buy more powerful and more expensive computers to keep up. If users are coming from around the globe, there is another problem—some users who are near the server will get fast responses, whereas users farther away will experience some lag. The following diagram illustrates a single, monolithic code base building to a single artifact:

What happens if the computer running this application has a fault, a power outage, or is hacked? The answer is that the entire system goes down entirely. For these reasons, businesses and applications have become more and more distributed. Distributed systems typically fall into one of several basic architectures: client–server, three-tier, n-tier or peer-to-peer. Blockchain systems are typically peer-to-peer, so that is what we will discuss here.

The advantages of a distributed system are many, and they are as follows:

  • Resiliency: If part of the system fails, the entire system does not fail
  • Redundancy: Each part of the system can be built to have backups so that if it fails another copy can be used instead, sometimes instantly
  • Parallelism: Work can be divided up efficiently so that many inexpensive computers can be used instead of a single (very expensive) fast computer