What is Fault Tolerance?

Fault tolerance is a system’s ability to resume operation when faced with interruption or component failure. The purpose for implementing a fault tolerant system within a company is to prevent disruptions arising from a single point of failure (SPOF) in order to safeguard high availability and business continuity. 

Fault tolerance is a process that can be built into a system upon inception, or implemented into a system after building. The entire fault tolerant design and execution depend on core technological processes, namely load balancing and failovers, to remove a SPOF. Successful fault tolerant systems will be able to transition between operations seamlessly and gracefully, ensuring that core processes execute under the most dire circumstances.

The Importance of Load Balancing and Failovers

Although what fault tolerance means between companies may vary, it is undeniable that at the core of fault tolerance lies load balancing and failovers, the practices responsible for mitigating the risk of SPOFs.

Load balancing is the method of network traffic distribution in which traffic is distributed appropriately across a pool of servers and resources. This workload distribution allows systems to optimize how they’re handling traffic, ensuring that high volume is handled and that stress is not placed upon a singular server. In addition, load balancing can partially combat network failures, temporarily mitigating slowdowns.

Failovers refers to the system’s ability to switch over to a backup system seamlessly and automatically in response to failure. Once components fail, the system will be switched to a backup system operating simultaneously with the primary system. In comparison to load balancing, failovers are only used during the most dire scenarios, transitioning systems until the primary system is back online.

Load balancing and failovers are best used within web applications, deploy redundant systems that make it hard for SPOFs to exist.

Fault Tolerant Operations

At its core, fault tolerance has two models that it abides by:

Normal Functioning 

Normal function refers to a situation when a fault tolerant system actually has a faulty component, but continues to operate normally. The system in question experiences no change in performance or metrics due to service interruption due in part to redundant components.

Graceful Degradation

When failure occurs, the impact a fault can have on a system depends on the severity of the failure. Certain fault tolerant systems will experience graceful degradation, where the severity of the fault is proportionate to the impact on performance. Like the saying “the bigger they are, the harder they fall”, the bigger and more severe a fault is, the harder a system will fall.

Fault tolerant operations do everything in their power to minimize the severity of a fault, relying on three core principles to guide development and mitigation efforts.

Principles of a Fault Tolerant System

Load balancing and failovers are just two pieces of the puzzle: what is the true fault tolerance meaning?

The primary goal of fault tolerant systems is to be self-sufficient, operating even under the most dire circumstances and failures.  To achieve this, fault tolerant systems operate under a set of principles that guide development, with these principles including:

Diversity

Although having identical backups is extremely useful in the case of a service interruption, diversifying alternative sources could mean well in the long run. Like a Swiss-army knife, fault tolerant systems should be equipped with all the backups and alternatives necessary, even if having them might be redundant for a bit.

Redundancy

Redundancy within a fault tolerant system is a means of removing a SPOF. Once a component fails, the system will be able to identify the fault and automatically replace it, without impeding operations. For example, in the case of hardware, a redundant array of independent disks (RAID) is usually put in place in the case of a system or hardware failure, letting users protect and access mirrored data without interruption.

Replication

In order to create the same instance of software as when a fault occurred, replication is extremely important. This would mean having the same version of software as the primary one, while having it run and operate as the primary one would. Function, tests, results- everything is identical to the primary system in the case that it would eventually replace the primary.

Fault Tolerance Components

Most, if not all, fault tolerant systems contain backup components that seamlessly replaced faulty components to ensure that a critical service is able to continue operating as a system fails, these components are:

Hardware

Hardware systems can be backed up by systems that are identical or equivalent to them. These backup systems run in tandem with the primary systems and mirror the operations. Think of the car analogy made earlier: a spare tire would be a good example of a replacement hardware system.

Software 

Software systems become fault tolerant once they’re backed up by other software. For example, having sensitive customer information backed up into a separate, isolated database and on entirely different computer systems ensures that related services continue operating in the case of a system failure. As mentioned above, failovers and having redundant components in case of system failures are of the utmost importance, with this philosophy applying heavily to software components in a fault tolerant system.

Power Sources

Power sources are able to become fault tolerant once a replacement power source is identified and ready. Having an alternative power source at the ready once a system experiences a loss of service is absolutely critical to uninterrupted operation.

High Availability vs Fault Tolerance 

Although fault tolerance and high availability are deeply rooted in each other, the two are not necessarily the same. 

Fault tolerance is a system designed to prevent outages and disruptions, deploying techniques to balance traffic and mitigate faults. High availability is a state of system, with a highly available environment providing strong uptime with minimal service interruption.

The fault tolerance vs high availability debate seems like a no-brainer, right? There is actually quite a bit of nuance to this decision, however. For instance, achieving fault tolerance requires specialized hardware and replicated systems, warranting a higher cost, though with a stronger quality of service. High availability, on the other hand, is a philosophy that developers used to guide their development efforts. Using resource sharing efforts, high availability would strive for 100% uptime, but could not guarantee it.

Choosing between fault tolerant computing and high availability depends on available resources and required uptime. For systems that need guaranteed uptime such as government and medical systems, fault tolerant design is the way to go. Organizations cognizant of costs and resource consumption will lean towards highly available systems, a great option for securing systems.

Fault Tolerance with Lyrid

At Lyrid, we’re enthusiastic about saving your applications and services from harmful events (while saving you time and money in the process). In order to address fault tolerance, we turn to our usage of microservices.

Through the use of microservices on Lyrid, you’ll be able to compartmentalize different apps and services that are running, meaning that if an application were to go down, this will be an isolated event and won’t impact your other services. This practice also translates to how we use Kubernetes; if a node were to stop, other machines would distribute the faulty node’s load and balance the traffic.

In securing your applications and services, we also offer disaster recovery and offsite backup machines, with regularly updated data.

Want to learn more about how we protect your system? Book a demo here!