NS3.40 60GHz WiGig (IEEE 802.11ad)

[A M]

I’m working on a network simulation project using ns-3.40, focused on modeling 60 GHz WiGig (IEEE 802.11ad) wireless links between two servers/racks in a Data Center Network (DCN).

The goal is to simulate how a wireless 60 GHz link could replace or supplement Cat6/fiber interconnects, and analyze performance metrics such as throughput, latency, and reliability under different traffic conditions.

Any Help is appreciated.

Kind regards Anthony

[Tommaso Pecorella]

Without doing simulations, and based purely on speculative ideas (i.e., I might be completely wrong), your data center will become a gigantic microwave oven, and this is not exactly a fantastic idea.

I'm not saying that you shall not do the research - just be ready to embrace the conclusions, and turn your presentations to "this is a very bad idea, and we demonstrate why".

Anyway, some tips:

- Use 3.47 if possible, 3.40 is outdated.

- Carefully consider the room and building characteristics and layout, and eventually update the channel models.

On the 2nd point, consider that many channel models are simplified, and do not fully apply to closed environments where you have tight spaces and a lot of metal, which is a typical layout of a datacenter. 

[A M]

Thanks Tommaso — you’re absolutely right about the propagation and reflection challenges in metallic environments, but my project assumes idealized LOS and does not account for metallic multipath.
I’m using ns-3.40 because the WiGig module (from the ns-3 WiGig repo) was last maintained for that version — later releases refactored the Wi-Fi stack and broke compatibility.
My simulation focuses on short, directional 60 GHz rack-to-rack links, similar to experimental DCN prototypes already demonstrated to offload elephant flows or supplement wired interconnects

[Tommaso Pecorella]

Sounds a reasonable choice - at least for a preliminary research.

Some more suggestions. Start by drawing a careful diagram of the network setup, i.e., who's going to decide what link to use.

If the application should be oblivious of the fact that you have 2 links, you'll need a shim layer to redirect the flows, which might be at L4 (TCP or QUIC), or lower, down to the NetDevice.

About multipath L4 (TCP or QUIC), there's Multipath TCP and Multipath QUICK (still a draft afaik), but I don't know how hard is it to integrate them into ns-3.

However, it could be a good idea to consider them, as they might allow you to avoid to reinvent the wheel.