Special vs Custom vs Standard? Choosing the Building Blocks-8
Continuing from last week, let us dig deeper in the IP Optics building block in reference to the Metro Networks.
If you leave aside some boundary cases and some parts of the world geography where the fiber infra is in dire state, the metro networks are pretty simple in design. The loss budget it limited, the distance is limited and the core locations are a few Data Centers / Mobile Core / Fixed Network Core or some cloud provider. The traffic patterns are hybrid, with majority of the traffic being hub and star and a minority that is any to any.
The network can be created with two layers,
- one layer with a lot of grooming and switching to handle the any to any traffic. This will be a lot of short hop, point to point connections
- second layer with a lot of optical bypass, that supports the hub and spoke traffic for different hubs, many of which can be collocated
The truth is that the main challenge that these networks face is today is just one: Scale! They cannot scale fast enough to meet the demand. They are going to face two more challenges in the coming future with 5G and IoT etc. namely,
- Latency
- Dynamism or flexibility
We need to reduce layers and simplify the infrastructure to make sure that the networks can scale fast, cost effectively and also in a way that the flexibility and latency requirements can be met.
If we look for the people who have achieved all these three today, we can surely consider the web-scale guys. So let us understand what they have done to achieve all this in very short term:
- Remove Custom: Welcome standard or White Box model: COTS for servers.
- Simplify IP: use switches and a cut down version of IP.
- Simplify Architecture: Leaf-Spine for inside scaling and DCI for outside scaling.
- Open Software such as ONOS for control.
- Simplify Operations: IT guys instead of telecom guys, stress on program-ability and use cases. No one needs a God Box.
Can we learn something from them? Or rather have we learned something from them? or Rather are they pushing for some technology that can change the face of metro architectures?
The answer is “Yes”, and we will discuss that next week.
So stay tuned.