The first post “Why the future of telecom will not be its past”, sparked a great deal of interest, primarily due to the errors it contained, some of which are big and basic, some of which are subtle and small. Please read this post first before you continue, and see how many of the errors you can spot.
It took Edison 1,000 attempts to make the lightbulb. Some people described this as 1,000 failures. When asked about this, he responded as outlined below. In his own words…
When a reporter asked, “How did it feel to fail 1,000 times?” Edison replied, “I didn’t fail 1,000 times. The light bulb was an invention with 1,000 steps.” [“They did not give up”]
To make sure you do not accidently see the errors, find a photograph that will hopefully fill your screen above the fold.
The first and the largest, most basic error. The area of a circle is , not pi * radius, as in the original post. This is the most embarrassing mistake of them all.
The next error is the assumption that the speed of light through glass is the same as the speed of light through the air. This is not true. The speed of light in glass is approximately one third slower than in air.
Then there are a number of additional points that are not taken into consideration at this moment:
the optical fiber never follows a straight line between source and destination.
latency in the radio link (much smaller in 5G than 4G)
latency in transmission overhead
latency in core network overhead
any level of redundancy, high availability
Thank you to Daniel for many of these additional model steps. I also added the last one myself. These can also be considered errors, if omission or simplification is classified as an error. We have disregarded this in our simple introduction since we still have 997 more steps left before we equal Edison’s performance, and we do not want to rush perfection.
Therefore for the same performance of 1ms transit to the data center and 1ms back (2ms roundtrip), the solution starts to look like the image below
Speed of light through glass
Number of datacenters with 2ms roundtrip
To consider if we bring closer so 1ms roundtrip looks like this:
Number of datacenters with 1ms roundtrip
And in this simple model, the conclusion is outlined below:
Range of datacenters needed
In this simple modeling for latency between 1ms and 2ms from source to destination and back, the number of datacenters required globally will be greater than 1,746 and less than 6,981.
Please feel free to comment on any new errors, or any other matter.
Please see post number three in this series, “5G VR — The NEXT Virtual Reality”, which introduces what use cases might need such performance and why.
Virtual reality is the “canary in the coalmine” of the future.
For more on Software Defined Infrastructure and hyperscale see here.