Marketers are liars.
You probably know this; or at least want to believe it.
Your network optimization company could be lying to you as well. Maybe just misleading you.
It’s unfortunate there is misinformation about how much you can improve your network and download speeds.
8X. 10X. 20X.
What do these numbers relate to? One can only assume the “X” refers to a multiplier. The problem is whether these improvement numbers are accurate.
CapitalOne has determined speed is its biggest concern. I am sure they realize they will not be able to increase that speed 20X. Especially when we are discussing network latency. Knocking down the Great Firewall of China will only get you about a 150 ms improvement in latency. And this is a known hindrance to network performance.
I will discuss what we can expect from network latency and how we measure network latency. Also, how we should report network latency improvements.
What is Current Network Latency?
When discussing network latency, anything over 200 ms is undesirable.
To begin, we can take a look at the Datapath.io network latency map below. The map is set for Amazon Web Services (AWS) global latency figures.
As we can see, latency is well below 200 ms for the AWS default route. To determine possible improvements, we need to understand the lowest possible latency. The limiting factor is the law of physics. The speed of light.
The Speed of Light Limits Network Latency
The lowest possible latency is going to be the speed of light. This is about 60 ms for the distance between Frankfurt and Washington D.C.. In our map above, the latency is about 109 ms, and the distance is about 6,500 KM.
Given this measurement, the greatest improvement we can see is a 45% improvement. That is less than a 2X improvement. This improvement is only realized if data is moving through the fastest possible route. Globally, improvements can be greater. In the instance of Europe to the United States there is the fastest fiber optic connection.
Given roughly the same geographical distance, we will look at a second location. That is Frankfurt to India. We can see below that latency to India is 179 ms. At optimal latency improvements of 60 ms, the improvement can only be 66.5%. Just under a 3X improvement.
Based on the global latency above, and our calculations, we can only expect a 60% improvement. That is with the fastest routes. Given this, we argue most networks will not allow for an improvement of 8X, 10X, or 20X.
How Should We Measure Network Latency?
Although expressing numbers in high multipliers is eye catching, it is unnecessary. To measure network latency improvements, we need to report accurately. In industries where latency is important, any improvement will be considered. Take the AdTech industry for example.
In a latency sensitive industry such as AdTech, an AdTech company cannot experience higher than 100 ms in latency. The reason is the RTB process will not consider an advertising bid which arrives in higher than 100 ms.
Thus, the improvements we see will be expressed in percentages. They are not in “X” improvements. The main reason is most AdTech companies already need to be operating under 100 ms. The greatest possible improvement cannot be a multiple.
Before making latency improvement measurements, we need to understand realistic starting latency measurements. Global latency of over 250 ms means your company is not utilizing a content delivery network (CDN).
If you are experiencing latencies of over 500 ms, then you are most likely sending traffic into the most rural areas of China, not utilizing a CDN, and have any number of potential barriers slowing your network.
The lie is how we report network latency improvements. Not how we measure network latency. Network improvements are important, essential, and well worth the investment. We need to manage expectations, and 20X is not realistic.
But hey, maybe we’re lying to you. After all, we are Marketers.
We discuss in detail how to determine latency benefits to your business in our eCommerce Whitepaper. Download your copy below.