GUEST OPINION: When it comes to CDNs, increasingly there is safety in numbers.
For internet-facing or internet-dependent organisations, redundancy and route diversity have long been desirable characteristics of service access and delivery.
In network-connected environments generally, there are many possible paths for traffic to take between a user and the data or applications they’re seeking to access. Some of those paths are more performant, reliable and lower-latency than others.
Even before the days of the internet, it was considered advantageous to be able to reduce or optimise the number of network hops between a user and the system or data they wanted. An efficient interaction promoted a better user experience.
Content delivery networks further optimised this by caching content geographically closer to user populations, shortening the distance between users and what they were trying to access, which, in turn, improved the performance and user experience associated with internet-based services, such as websites and apps.
One of the enduring challenges faced by web-based service operators is that user expectations for continuous improvements and consistent performance have not plateaued over time. As more of the world relies on the internet backbone to function, it has brought into view the way that web services and service delivery is architected, and the extent that they are resilient to faults or unplanned outages.
Entire disciplines such as site reliability engineering have come to the fore in helping to address some of these issues – but so have better- and best-practice architectural blueprints that promise to produce more available, reliable and stable performance of websites and web-based applications.
Redundancy – once a common conversation in the on-premises world when looking to achieve highly available applications or services – has renewed relevancy to this discussion.
Specifically, the use of multiple content delivery networks – long seen as a way to improve the quality of experience (QoE), and create an adaptable infrastructure to handle local bandwidth issues – is gaining traction and adoption among Australian businesses as a way to further improve resilience and availability, and as a hedge against degradations and outages, such that if the primary CDN encounters issues, the content can still be served via a different CDN, so that continuity of the user experience is assured.
Tracing the appetite for a multi-CDN model
Traditionally, organisations involved in delivering web-based experiences have typically prioritised a single-vendor CDN solution that is super reliable and that can consistently deliver content safely to their customers. Some of the reasons for choosing a single vendor
did (and still do) come down to convenience and cost. Adding providers adds cost and potentially complexity to the experience delivery mechanism.
But what is increasingly being recognised is that downtime in a world that prioritises always-on and always-available web-based services comes at a cost as well.
In both regulated and non-regulated industries in Australia, maintaining service, user experience and delivery offers competitive differentiation today, but it’s also fast becoming table stakes for all organisations.
This is driving more organisations to adopt a multi-vendor, multi-CDN strategy and approach as a design best practice to reduce single-vendor risk and to keep traffic flowing in the various link, node, and site incident scenarios that might otherwise impact availability and performance.
By distributing traffic across multiple CDNs, organisations can more seamlessly reroute around outages or regional bottlenecks, maintaining uptime while reducing the risk and cost of downtime. Intelligent traffic allocation ensures users are always served through the fastest available path, delivering consistently high-performance experiences worldwide.
The right model
As more organisations go down the multi-CDN path, we often find they often lack a strategy for how to measure their CDNs’ performance, and how to allocate traffic across different CDNs.
There are a number of different models for operating and orchestrating multi-CDN environments. Adopting a platform that is designed from the beginning to operate in a multi-CDN model, that can orchestrate traffic and abstract away complexity is important. And when combined with an effective partnership approach, organisations are able to overcome challenges associated with a multi-CDN model while yielding the enormous performance, availability and resiliency benefits that such a model promises.
While there are several popular design methods for a multi-CDN approach, traffic splitting – for example, using DNS-based steering or Layer 7 traffic management techniques – is one of the most common. Determining the appropriate traffic split is important. It may be that traffic is split evenly – 50-50 – between CDNs so they effectively share the load. Alternatively, traffic splitting portions may be more dynamic, based on ambient internet and network conditions, such that one vendor appears to be having issues, traffic can be routed to a different CDN and around the problem.