A recent article written by Sebastien Anthony at ExtremeTech talks about a massive project that is underway to improve the network latency between the UK and Japan from 230ms to 170ms. The project involves the laying of three fiber optic cables that traverse the Northwest Passage which runs through the Canadian Arctic Archipelagoand and the north coast of Scandinavia and Russia and leveraging a ton of branches to make a connection between the UK and Japan. The cost of this project is expected to be between $600M and $1.5B.
I can't help but to wonder if other options were investigated before deciding to invest in this massive effort for what turns out to be a latency improvement of only 60ms. The end result is that a one-way packet delivery from the UK to Japan will still take 170ms. Obviously an improvement over 230ms, but still very high when you compare the potential alternative.
So what is the alternative? It seems like this would have been a tremendous opportunity to implement a WAN optimization solution. WAN optimization technology, like that offered from Riverbed, mitigates the impact on high latency by reducing the amount of data and associated round trips required by the applications that are delivered over the network connection.
Techniques such as data deduplication and protocol optimization can often improve performance 10 times, 50 times, and up to 100 times or more in the most severe of high latency environments. Recent breakthroughs from Riverbed such as Granite Virtual Edge Server Infrastructure take on high latency constraints head-on. There are various flavors of WAN optimization from appliances to software to cloud-based services.
The good news is that it is not too late! The incremental improvement from 230ms to 170ms that the new bandwidth infrastructure provides can be looked at as the first step. Add WAN optimization on both ends of this 170ms connection and there is the potential to reduce the impact of the latency so it feels like both ends of the connection are in the same building.
This is just another example of how bandwidth upgrades is not the end solution.