Steve Jobs' Near/Far WiFi Problem, Explained (Video)
Steve Jobs introduced the iPhone 4 yesterday. You may have heard about this snazzy little device and you may have also heard about the problems Mr. Jobs had in demonstrating it. Mr. Jobs blamed the problem on WiFi, and as best I can tell he was right. I whipped up a quick video to explain what likely happened.
It's my first video blog post, so be gentle. Thanks to Rough & Tumble Films in Los Angeles for providing the space and Nick Robinson (@nickrob on Twitter) for helping me out with the shoot.
It's my first video blog post, so be gentle. Thanks to Rough & Tumble Films in Los Angeles for providing the space and Nick Robinson (@nickrob on Twitter) for helping me out with the shoot.
hi Ben,
ReplyDeleteI enjoyed your explanation of the Wi-Fi problem on Steve Jobs recent iPhone release. However, I think the problem was more likely due to excessive numbers of 802.11 management frames being transmitted. I doubt if there were 300 or so Wi-Fi hot spots (i.e. Mi-Fi devices) present in the auditorium, I suspect there were 300 STAs busily transmitting probe request frames to a few APs in order to gain network access. Typically, an un-associated STA will transmit a few probe requests frames a second and if you then multiply this by several hundred client STAs trying to do the same thing, then you have a very busy wireless medium and more importantly the collision probability will be high, probably so high as to render any form of reliable communication almost impossible. Furthermore, even if you did manage to become successfully associated with an AP, the capacity available will be so small that it will probably cause any http or tcp/ip session to time out.
I have to disagree. The Apple Airport client in OS X, and the Windows clients for XP SP3, Vista and 7 all disable Probe Requests after a station is associated.
ReplyDeleteIt is possible that management frames could have been a problem due to excessive Beacons, but that would go back to the problem of having too many APs on one channel.
Interesting approach as for the issue understanding, considering that most or all of the APs could've been in the same power output range; as for the solution, fewer APs or one AP per channel is an obvious solution but in Job's case, I think they could've been one step ahead and consider your visitors\attendees would bring their own APs, so, Job's people should've set up an AP with stronger signal, by simply attaching a stronger antenna (i.e. 5dbi) because, stronger antenna = stronger signal = better reception = steadier connections, but maybe they just didn't see it coming....
ReplyDeleteHello from 15 years later! I hope you're still around, Ben.
ReplyDeleteI'm curious what you think about the possible impact of the near/far problem in more modern Wi-Fi? Specifically 802.11AC and AX, but on 2.4GHz, not 5GHz.
I ask because I perform wireless network performance characterization for Wi-Fi and 3GPP networks. Recently, we were trying to trigger the near/far problem with a Wi-Fi6 AP on 2.4GHz only, a Wi-Fi5 "far client" and a Wi-Fi6 "near client" streaming a constant 6mbps UDP via iPerf3 (and an initial test at 12mbps). We found that in the 12mbps near-client test, after just a few hundred meters (~350m exactly across 3 tests) the far clients completely lost the ability to send data on the network and eventually disassociated. At 6mbps, we still noticed frequent instances where iPerf throughput would drop to 0, but intermittently we'd still see nominal throughput until ~1KM, when the far client disassociated completely.
We initially took this as correct replication of the near/far problem and felt it was pretty disastrous news for a wide-ranging Wi-Fi network with a single AP. After further research though, I suspect we really actually triggered the "hidden node" problem instead. The more I research 'near/far' specific to Wi-Fi's CSMA (as opposed to CDMA networks, which the Wikipedia article specifically references), I only really find old articles, like this one, or another I found from 2003. Up until I saw yours here, I had convinced myself that near/far is NOT a factor for Wi-Fi at all.
So, ultimately, I'm curious how accurate you feel this video is 15 years later (still on 2.4GHz, 20MHz channel, with Wi-Fi5 devices on a Wi-Fi6 AP). I'm also very curious to know if you think OFDMA in Wi-Fi6 would negate either near/far or hidden node.
Signed,
A Network Engineer in Need
P.S., I'll keep checking this quite regularly, but feel free to also email me at william.a.dell@nasa.gov