What Switch Requirements Does 802.11ac Wave 2 Bring?
I have been speaking with many customers recently who are going through a process of upgrading their edge networks. The most common topic is that current switches are old and are coming to EoL. And a question that keeps coming up is whether or not they require 2.5 Gbit/s and 5 Gbit/s speeds over Ethernet because of the advancements in the 802.11ac wave 2 technologies (i.e. MU-MIMO).
Most customers buy networks on a 3-to-5-year timeframe, so regardless of whether they need such high speeds now, they need to forecast network demands at the end of the time frame.
So firstly what is this technology, and what does it do? The IEEE 802.3-2015 amendment defines enhancements to the Ethernet Media Access Control (MAC), Physical layer, and management objects to produce a standard for 2.5 Gbit/s and 5 Gbit/s speeds over Ethernet twisted pair copper wire.
So coming back to the original question, whether or not they require 2.5 Gbit/s and 5 Gbit/s speeds over Ethernet. To enable us to understand this we would need to make some assumptions and model, based on those assumptions.
The specific question I received from the customer was “supporting a new installation of 802.11ac wave 2 technologies for the next 5 years, are we likely to require more than 1Gbps connectivity for each AP?”. Remember, 1 Gbps Ethernet is full-duplex which means it supports 1Gbps up and 1Gbps down (that’s quite a bit of capacity!).
This is a good question, but in order to answer this we need to calculate the bandwidth created by the specified clients connecting to the network with specific conditions the clients are performing to. Of course it is possible to collate this information from existing clients. However, the customer had specified that this was a network for both now and the future, therefore we would need to make some assumptions.
These assumptions represent a worst-case scenario, so the real-world needs are likely to be less than the model. This model is based on aggressive assumptions as a “stress test”, this is so we can plan on a worst case scenario.
- In 5 years 100% of the clients connecting to the network will support 802.11ac
- The clients connecting will support a mix of spatial streams
- 50% will support 4 streams
- 40% will support 3 streams
- 10% will support 2 streams
- Everything will be using the fastest possible (MCS9) rates
- The clients connecting will support a mix of channel widths
- 20% will support 20 MHz
- 50% will support 40 MHz
- 30% will support 80 MHz
- Every client connecting to the AP will support MU-MIMO, this means that the AP always does 4 streams. It’s important to note that in the real world, MU-MIMO probably reduces the speed of each stream, but in this case, we assume that the AP is capable of 4 streams at MCS 9.
- We specified that we aggregate 40 frames into every A-MPDU
- No contention
With these assumptions now made we need a tool take these inputs and formulate the required outputs (our AP Ethernet bandwidth).
So I used Matthew’s tool to input the assumption data and the following is the output we came to.
- Not all clients will be 802.11ac in 5 years, there will always be legacy to support (i.e. 802.11n clients on 2.4GHz and 5GHz)
- It’s hard to forecast the spatial stream mix, but 4-stream devices have been relatively slow to market and are unlikely to represent 50% of devices, especially with the move to mobile.
- MCS 9 is a 256-QAM rate, and it is hard to sustain. Most clients will not be so close to the AP that this is sustained
- It’s also hard to forecast the channel mix, but with the limited number of channels it seems unlikely that large numbers of transmissions in high density environments will be at 80 MHz.
- 100% adoption of multi-user MIMO is extremely aggressive, and specifying that it supports 4 simultaneous streams outstrips the state of the art
Even our constructed worst-case scenario doesn’t require more than 1 Gbps to the AP. Of course, when you look at areas of high AP density, multi-gigabit (whether 2.5, 5, or 10G) uplinks from the edge have a role to play in your networking strategy.
Once I explained all this to a customer, they had trouble justifying the cost of the switch upgrade just to service high-density Wi-Fi. In my case this really helped the customer decide what they needed to do moving forwards with their edge upgrade.