Special vs Custom vs Standard? Choosing the Building Blocks-16

Continuing from last week, let us look at one of the problems once the Data Centers move to the edge and get distributed.

The economies of connectivity will change drastically. Implementing a spine leaf architecture within a Datacenter is within the control of the owner. They can lay as many fibers as they want and they can connect each spine with each leaf on a separate fibre (pair) using 100G or more. However when this spine leaf is spread over a geography, the Datacenter is “Virtually” a large Datacenter, but in reality is in fact the leafs are located at separate geographical locations compared to the spines.

We now have a limited supply of fiber, the link budget is constrained and the reliability of the fiber is not guaranteed. We will need to adopt a different architecture on the transport, DWDM for sure, programmable for sure, and multi-hop for sure, it may also include some hubbing locations etc.

The cost per GB now changes a lot as the power, space and connectivity costs are hugely different from a hyper-scale data center. I do not believe that the webscale organizations can solve this problem on their own. They need the carriers. So look out for more alliances. The Carriers and webscale have been fighting each other for some time now, so it may be difficult to collaborate and it will take time for the right commercial models to emerge. But make no mistake the alliances are the way forward.

Next week we will pick up another aspect of the distributed connectivity. So stay tuned.