Creating software that operates seamlessly in a distributed computing environment is a complex endeavor. The task becomes even more challenging when developers fall into the trap of believing the "Fallacies of Distributed Computing." These fallacies are a set of assumptions about networked computing that, while seemingly reasonable, can lead to significant failures in software development1. Let's delve into these fallacies through three real-world examples and discuss how understanding them can improve the process of creating digital software products.
Also known as Fallacies of Networked Computing, the Fallacies are a list of conjectures (or beliefs) about distributed computing, which can lead to failures in software development. The assumptions are:
- The network is reliable
- Latency is zero
- Bandwidth is infinite
- The network is secure
- Topology doesn't change
- There is one administrator
- Transport cost is zero
- The network is homogeneous
The first four items were listed by Bill Joy and Tom Lyon around 1991 and first classified by James Gosling as the "Fallacies of Networked Computing". L. Peter Deutsch added the 5th, 6th and 7th fallacies. In the late 90's Gosling added the 8th fallacy.
The group was inspired by what was happening at the time inside Sun Microsystems.
These fallacies should be considered carefully when designing code which is resilient; assuming any of these fallacies can lead to flawed logic which fails to deal with the realities and complexities of distributed systems.
Examples
The Network is Reliable
One of the most common fallacies is assuming that the network is always reliable. In reality, networks can and do fail, leading to lost or delayed data transmission. An example of this is the failure of a messaging app due to network issues. If the app is designed assuming constant network reliability, it might not have proper error handling for network failures, leading to lost messages or app crashes. By acknowledging that networks can be unreliable, developers can build software with robust error handling and retry mechanisms, resulting in more reliable and user-friendly applications.
Latency is Zero
Another fallacy is that network latency is zero, meaning data is transmitted instantaneously. This is rarely the case, especially in globally distributed systems. For instance, a video conferencing app may suffer from latency issues, leading to poor call quality if not properly managed. Developers need to account for network latency and adopt techniques like data compression, caching, or using Content Delivery Networks (CDNs) to ensure smooth data transmission and improve the user experience.
The Network is Secure
Assuming that the network is always secure is a dangerous fallacy. In reality, networks are often susceptible to various threats, including data breaches and cyberattacks. For example, an e-commerce platform might be targeted by cybercriminals aiming to steal sensitive customer information. By accepting that network security is not a given, developers can implement proper security measures such as encryption, two-factor authentication, and regular security audits to protect the software and its users.
The Fallacies of Distributed Computing are a reminder of the complexities and realities of building software for distributed systems. By challenging these assumptions, developers can create robust, efficient, and secure software products that deliver a superior user experience. Understanding these fallacies is not just about preventing software failures—it's about cultivating a mindset that appreciates the complexities of distributed computing and uses that knowledge to build better digital products.