Skip to main content

How Viable Is SDN for Large-Scale Networks?

·2 mins

SDN is fast, yes. But, how viable is it to replace the existing traditional networking model? Consider the current traditional networking model in all it’s glory. There’s more than 12 devices in my flat with 6 people.

The fact that we got till here is an engineering marvel to be very honest. Handling 12 devices per flat for an apartment complex of 60 people begins to give us an idea of what kind of traffic one building may provide us with.

Taking only my flat into consideration, if I were to start a download for a game that is approximately 50GB in size, a reasonable expectation on my end would be for it to finish in 15 minutes. Too less? Okay. Let’s consider it to be around 20 minutes.

The equation should look something like this.

download_size(MB) = network_speed(MB/s) * time(s)

I don’t trust my math, so I’ll leave it to a python interface instead.

# 50 gb to mb
DOWNLOAD_SIZE: int = 50 * 1024

# Take 5 MB/s despite network providers advertising in Mbps

# Calculate the time taken in seconds to download the game.

# Human-readable statistics
time_in_minutes: int = int(time/60)
network_speed_in_bits: int = NETWORK_SPEED * 8

# Returns how long it takes for the download
return f"Takes {network_speed_in_bits} Mbps for {time_in_minutes} minutes"
Results: Takes 320 Mbps for 21 minutes

In a lower-bound scenario where there’s 6 people wanting to stream high quality video content and download games, the required network speed levels hit 1920 Mbps, or 1.8Gbps.

Of course, surfing through the internet and authenticating yourself on a website isn’t a bandwidth-intensive task, but it goes to show what kind of speed should the network be able to support.

And the question is, whether SDN can? Some of the vital questions that need to be answered to properly evaluate the potential advantages or shortcomings of SDN are,

  1. Evaluation of SDN in large-scale networks and how much bandwidth can be provided to each node.
  2. Evaluation of latency in a significantly expansive network architecture.

Based on the above two statistics, a question that needs to be answered is whether SDN can cater to the general public’s usage of the internet or whether it’s a specialized network made for campus and enterprise-grade on-prem networks.

If I do get the time, I’ll make a note of it here, or I’ll just link a bunch of research papers that do the same.