Choosing a CDN - Part II: How To Evaluate CDNs

Last week’s blog post on choosing a CDN walked through what really matters to the industry's clients. For those who haven’t read it, ultimately, it came to the conclusion that what customers really want is a CDN that delivers their content well, 24/7. But how can you figure out which CDN that is? I think the 3 metrics that create the best holistic overview are: Network Availability, Latency and Throughput.

CDN Performance

Network Availability

CDN availability is the time the network is working properly divided by the total amount of time in the period. In other words, is the delivery system working and transmitting the requested data? Most customers need their services to be running properly as close to 100% of the time as possible, because their end-users could attempt to load content at any moment. Since users can often be anywhere in the world, a network that reaches all of them is important for availability too.


CDN latency is the amount of time between when a user makes a request and when content starts to be delivered. The shorter, the better. A normal amount of latency for websites visited via computers is frequently in the 100ms range. It was once reported that at Amazon, every 100ms of delay cost them 1% of their sales. Whether a site is an ecommerce one or not, engaging users is of tremendous importance to almost everyone. Web pages that load slowly are at a huge disadvantage compared to ones that load quickly.

CDN latency can also be related to global reach, as end-users can be spread out among large geographic landscapes. Many CDNs discuss their points of presence (POP). In general, more POP means that information has a greater global reach but, the location of these POP is very important. The closer a user is to a POP, the closer he or she is to the requested content, and the lower his or her latency will be.


CDN throughput is related to a network’s ability to support data transfer (i.e. how much load it can sustain). If a CDNs network can’t transfer large amounts of bandwidth at the same time, it can get overloaded, which can lead to high latency and a bad user experience. Due to technicalities relating to internet protocols, latency can also drastically affect throughput. You can learn more about the technical details behind that issue here.

Different applications have different needs, but all customers must make sure that they're allocated enough throughput so that they don't have a bottleneck. Again, the importance of this comes back to user engagement, since servers operating without excess capacity can stop your customers from getting to what they want.

Anything less than 24/7 content distribution from a CDN could cost your company business/money. While no system is perfect, when choosing a CDN understand that the best one’s will consistently show very high levels of network availability and throughput, as well as low latency.

Feel free to follow the links to some real-time test data, provided by Cedexis, or run tests yourself.

If you’re interested in boosting your CDNs performance, click here to learn more about how Peer5 can boost your site’s capabilities.