Will 802.11ac fire the hospital?

Today, the industry seems to be plagued by the new 802.11ac standard.

Now 802.11 acWave 1 has been launched, manufacturers promise to achieve a transmission rate of Gbps in the 5 GHz band, each access point can achieve a connection rate of up to Gbps. However, all industries are concerned about the possible bottlenecks of AP backhaul networks. In other words, is a single Gigabit Ethernet uplink sufficient to ensure access points?

The answer is simple: "Yes", not only for Wave 1, but also for 11 acWave 2. The reason is as follows:

In reality, there will be no theoretical maximum values.

Although 11 acWave 1 is expected to reach the theoretical rate of Gbps aggregation, it is difficult to reach the theoretical maximum in the real world, and it will not.

1.75 Gbps is the data rate. However, the real TCP throughput (that is, the user experience) has always been around 50% of the data rate. With the 11n/ac frame aggregation and other improvements, even in the best environment, it can only achieve about 65% (usually only in a single customer test ). Therefore, for the purpose of discussion, 65% of the theoretical value is possible, that is, about 1.15Gbps.

Therefore, assume that:

Only 3x3: 3 client devices, one in the 2.4GHz band and one in the 5GHz band

Excellent RF conditions without interference from neighbors and other RF Frequencies

There are TCP applications that can maintain Mbps

100% upstream or downstream TCP applications

So it may be possible to achieve a gigabit return, at least people argue.

But this will not happen in the real world.

The client device combination does not support maximum performance.

If the network is composed of all client devices that support 80 MHz channels (in the 5 GHz band) and three-space streams, there may be a probability that the stars are arranged in a column, although the probability is very small.

Back to reality:

1. There will be some single-stream client devices, such as mobile phones and tablets.

2. There will be some dual-stream client devices, such as tablets and most laptops.

3. There will be some 11a/g/n devices that do not support the maximum 11ac value.

4. Some customers may have a service area more than 3 meters away from the access point, so it has a low data rate.

Therefore, as long as there are any such client devices on the Network (and indeed !), Then you can kiss the Gigabit saturation nightmare. Each low-performance client on the network will reduce the average call efficiency, so that Gigabit pressure conditions no longer exist.

Don't forget that Ethernet is full-duplex

When comparing the Wi-Fi network speed and the Ethernet speed, remember that the Wi-Fi network is half duplex. All call (airtime) transmission is shared by the uplink and downlink. Therefore, if there is a theoretical maximum channel capacity, it must be divided into uplink and downlink. On the contrary, Ethernet is full-duplex, with 1 Gbps uplink and 1 Gbps downlink synchronized. Therefore, to put pressure on this Gigabit link, you have to push all upstream or downstream traffic of all Wi-Fi network clients. Similarly, in reality, this will not happen.

Application software will not put pressure on 1 Gbps return links

Taking into account the performance limitations of client devices, a small number of client applications and services can generate sudden, consistent load requirements higher than 700Mbps. However, this is not the potential of a single client device, but the potential of all client devices that share incoming and outgoing traffic (airtime) transmission.

High density does not put pressure on 1 Gbps

At first glance, high-density networks seem to bring about Gigabit pressure, making it easier to reach the maximum network. However, in this case, high-density scenarios are more likely to involve single-stream mobile devices that do not support the maximum protocol value, resulting in retry and non-data-intensive call (airtime) transmission pressure, this reduces the overall network potential.

Today, most networks cannot deliver commitments.

How many networks can provide over 1 Gbps WAN links and deliver network-based services/applications? Today, it can be delivered through cloud computing, and most client-based applications use cloud computing services.

Local LAN applications/servers are more likely to continuously process 1 Gbps. Is the application at a rate higher than 1 Gbps in a specific direction and runs independently without any other client devices generating traffic? The answer is self-evident. No.

Cost is always king

From a business perspective, it is hard to believe that someone is willing to spend money on reaching the 10GbE edge for all access points, and no one is willing to spend money on Cat7 cabling of a higher level (to be honest, Cat 6 may be more reasonable ). It is clear that the cost of deploying multiple copper cables for each access point with Link aggregation is too high. As long as you show the budget owner the possibility that a single low-cost 1 Gbps link may encounter saturation in reality, the budget owner will reject this decision without hesitation. Therefore, if you cannot get through the technology, it is always useful in terms of cost.

What about 802.11 acWave 2?

All indications are that 11 acWave 2 AP is a three-stream (still) or more likely 4x4: 4 stream (5 GHz up to 1733 Mbps ). These APs also support a MHz channel with a higher data rate. Therefore, the reason for Wave 2 Gigabit return is that the 160 MHz channel is most suitable for small office environments and it is impractical to include it into enterprise-level products. In addition, most enterprise client devices are unlikely to support MHz Wi-Fi network channels.

4th streaming does not change the actual throughput pressure

Considering all previous discussions on client composition, application requirements, return issues, and high density, the extra space flow of the Access Point has almost or no effect on the return link. Even if there is, almost no client can support four spatial streams first. The total throughput of each access point will continue to be limited by low-and medium-performance clients. Even high-performance clients are difficult to generate one-way TCP traffic of nearly 1 Gbps.

Multi-user MIMO does not increase the maximum return pressure

Now you may want the ability of a MU-MIMO or access point to communicate with multiple clients at the same time, which may change everything. However, no.

Without a doubt, if there are many single-stream client devices, and the downstream traffic is the majority of the case, the MU-MIMO will increase the call (airtime) transmission efficiency. However, the access point still has only four spatial streams, and not each transmission uses MU-MIMO. In many cases, MU-MIMO transmission is only used on two single-stream clients, which is far from the Gigabit upper limit.

Everyone has neighbors.

Wi-Fi network performance is almost always dependent on RF conditions. Although in a lab environment, the maximum data transmission of Wave2 can be close to the Gigabit ceiling, the problem is that the same high-performance network must share the call (airtime) with its neighbors. Looking forward to the future, there will still be a large number of 802.11n networks, which will inevitably lead to backward compatibility.

Don't worry about Gigabit again

The implication of this story is that the theoretical environment may put pressure on a single Gigabit return, but in the real world, client composition, RF environment, application requirements, and network infrastructure mean that high-performance full-duplex Gigabit link full load saturation is impossible. It is not true that you want to persuade the customer to upgrade the wired network based on the theoretical environment and arguments.