Help With High Latency Seen On WiFi LAN Traffic For AC3100 & AC88u - Updated Test Info & Findings

[sfx2000]

And FWIW - the medianet router/ap - it's an old Linksys WRT160NL N300 class with stock firmware

For the media players, just need to throw bits out there with a 100BaseT connection back to the Gigabit backbone - and they don't ding the main network

11n on 2.4GHz can stream 720P H.264 streams well in a decent network...

 

[sfx2000]

and just to cut off the pass - intranet bandwidth on that old Router/AP - more than enough...

The 160NL WAN side is still private - but I've got a speedtest mini inside the main LAN - so this is WAN/LAN performance across that link over N130 (11n/20Mhz) - internal streams on the medianet, so far, just fine...

 

[Wildcatz_1]

Thats interesting to see. I did reply to your earlier post with extensive info on the VLANs being deleted, further testing but for some reason that larger post (just posted it about 5 mins ago) is awaiting moderation ?

and just to cut off the pass - intranet bandwidth on that old Router/AP - more than enough...

The 160NL WAN side is still private - but I've got a speedtest mini inside the main LAN - so this is WAN/LAN performance across that link over N130 (11n/20Mhz) - internal streams on the medianet, so far, just fine...

View attachment 6124

 

[Wildcatz_1]

Splitting reply into 3 posts: Post 1 of 3

If I saw the 1 Gbit/s WiFi you are seeing then that would be great. Which AP/Router were you using for that test (your Network Utility screen grab) ?
Disabled all VLANs on the Cisco, in fact deleted all completely, so all clear there.

AP is hardwired to the Cisco 8 port with Ether running as a test to the Internet port of the AC3100 as well as throwing the uplink into port 4 as well. Same issue regardless of port.

I also opened up the 5Ghz on the AC88u (normally only running 2.4 as mentioned in my diag) to test to see if the issue could be correlated to AP vs router mode (AC3100 is in AP Mode, AC88u is in Wireless Router Mode). Still same latency and same link speed

So unless there is an issue with the way in which these Asus routers are handling WiFi to WiFi traffic.......then would lead me to think the same thing as you, loop issue or something related to the Mac. Would of course feel happier if the Link Speed in Network Utility showed the correct speed as at least then could rule out client completely although as we both said, with Airport and Wireshark confirming what each other saw, that should point to a misreported speed rather than actual impact (hopefully).

Some further data points here:

Pinging From Network Tools Within AC3100 GUI
In the AC3100, using GUI Network Tools and pinging Mac Mini 1 connected to it via WiFi
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max = 43.173/211.015/505.621 ms

In the AC3100, using GUI Network Tools and pinging Mac Mini 2 connected to it via WiFi
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max = 43.564/160.940/278.065 ms

In the AC3100, using GUI Network Tools and pinging google.com
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max = 10.073/14.926/35.739 ms

In the AC3100, using GUI Network Tools and pinging AC88u (hard wired together through the diag I shared earlier)
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max = 0.237/0.269/0.393 ms

 

[Wildcatz_1]

Post 2 of 3:

Now switching to the WiFi Clients (802.11N & 802.11AC) Pinging Out Via OS X Network Utility To Wired

Mac Mini #1 802.11 AC Client (connected to AC3100) Over WiFi to AC88U
10 packets transmitted, 10 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 2.463/2.905/3.141/0.220 ms

Mac Mini #1 802.11 AC Client (connected to AC3100) Over WiFi to AC3100
10 packets transmitted, 10 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 1.342/2.696/3.724/0.706 ms

Of course as expected, in other testing, wired to wired is super low, under 0.3 ms so no issues there.

Has me thinking that if WiFi to Wired is not bad (but certainly not great) and Wired to Wired (even hardwired to AC3100) is super fast as expected (the under 0.3 ms), then overall routing should be solid and would (I would think) reduce the possibility of a loop etc as the issue only occurs on WiFi to WiFi. This would then put me back to the thought of it being Asus router related and the way in which is is handling the WiFi to WiFi traffic. However with that logic out the way, I would then expect others to be seeing similar issues if the WiFi to WiFi traffic handling of these units was the culprit. Unless this loop issue is trapped in the routing table bound to the WiFi Mac address of course. I also still think the Mini needs to be removed from the equation too just in case something weird going on with that Link Speed + any power saving issues with the Mini networking. With the said, the mini networking itself is pretty locked down as most Mac's are so trying to make any change to the built in AC is going to be tough unless you look at potentially trying to manipulate the kext files.

 

[Wildcatz_1]

I pinged out to ATV 4th Gen (from the 802.11n Macbook Pro connect to the AC3100). On the ATV 4 the pings start high, 1st is over 500ms then every subsequent is under 5ms (like power saving being on initially as discussed but kicking off quickly). This could then point back to the minis as an issue.

The ATV 4 is of course running tvOS. I then had a friend bring over an iPhone and iPad Pro and there the pings were high UNLESS I hammered the client. In other words if I did something network intensive on them, such as Speedtest etc the ping times reduce (from about 85ms to about 4ms under heavy traffic). This behavior of course can correlate to power saving and only ramping up when network intensive tasks are used (ICMP not being network intensive of course).

Therefore I don't think I can rule out the Mac Mini's just yet as they remain high latency on WiFi period (even under heavy load) and still have not fully resolved the 11 Mbit/s potential issue. I think I am going to have to try picking up another (non Asus) AC router just for testing in exactly the same environment as well.

For testing of the loop issue, I could take out the Cisco 8 port on the main floor, run the AC3100 uplinked via the wall port to the patch panel, bypass the Cisco 26 port and run the link from the room on main floor direct into the AC88u. That way physical hops would be reduced. However, since I tested above with both clients connecting to the AC88U via WiFi and pinging each other with the same latency issue, not sure what this would gain me.

Thanks for the continued help here, much appreciated

 

[sfx2000]

If I saw the 1 Gbit/s WiFi you are seeing then that would be great. Which AP/Router were you using for that test (your Network Utility screen grab) ?
Disabled all VLANs on the Cisco, in fact deleted all completely, so all clear there.

Airport running as an AP, not Router - I have a real router to do those routing kind of things...

My best guess at the moment - the Asus devices might have some issues with client-client communications across the radios - shouldn't be, but your numbers and testing suggests much...

I'm willing to bet, cabled up, everything would be right, even if only one is cabled up - and then it's how the traffic flows across the SoC/Firmware.

 

[Wildcatz_1]

That is what I am leaning towards as well which is a pain as nothing I can do about it

Thanks for the continued help here, much appreciated

Airport running as an AP, not Router - I have a real router to do those routing kind of things...

My best guess at the moment - the Asus devices might have some issues with client-client communications across the radios - shouldn't be, but your numbers and testing suggests much...

I'm willing to bet, cabled up, everything would be right, even if only one is cabled up - and then it's how the traffic flows across the SoC/Firmware.

 

[Wildcatz_1]

Did some more testing today with some interesting results.
Pinged from an 802.11n client (Macbook Pro) to the Mac Mini #1 I saw the expected high latency, however I was able to improve it with some mixed results.


First test here is from the Macbook Pro to the Mac Mini #1 while it (Mac Mini 1) was running a speed test on speedtest.net. Note the decline in latency when running speediest (just a short sample here in Pic 1):

I then did a test from the same Macbook Pro while the same Mac Mini was doing a 6gb file copy to a NAS again notice the drop when copying started (Pic 2)
I then wanted to try AC to AC testing so test pinging from Mac Mini #2 to Mac Mini #1 while it (Mac Mini #1) was doing a speedtest again notice the drop off of latency (Pic 3)
I then repeated the copy to NAS test this time from Mac Mini #1 while Mac Mini #2 was pinging it. Again drop off was evident (Pic 4) and begins to paint a pattern (potential aggressive power saving on the NIC).

However it then got very interesting. I then tried copying the same file from Mac Mini #1 to Mac Mini #2 while Mac Mini #2 was pinging Mac Mini #1 (pic 5). Here you an see there was no drop off at all. I even reversed this using the other Mac Mini and again no drop-off !


So this has left me with a couple of observations and a question

1) Link Speed being incorrect on Network Utility definitely ’appears’ to be more a bug in OS X misrepresenting it as again it will accurately report 802.11n (450 Mbit/s) when using other routers (have not tried another AC router yet but intend to). Wireshark captures and even Airport stats seem to confirm 1300 Mbps (AC) TX rates as does the router in its wireless logs and client connection list. Could be an issue with Broadcom comms between the Mini & Asus as well of course, different router should help eliminate this one.

2) It would appear that aggressive power saving could be in use by OS X and the Broadcom drivers in these Mac’s which unfortunately cannot be tweaked. Certainly appears more aggressive on these Mini’s than other devices I tested with.

3) Still a potential issue with WiFi to WiFi traffic on the Asus routers, maybe even power saving issues on the routers too ?) but is looking like the Mini’s may be playing a bigger role in the issue.

4) Very interested in picking up an Apple Airport Extreme (like you have SFX) for testing as an AP only and use these Asus routers on the backend to see if I can overcome some of the issues seen.


now for the questions:

1) Any reason power save mode (if it is that) seemingly disables when a Mini is running a speedtest or a copy to a NAS BUT when copying to another mini (remember same model year and spec as each other) does not appear to disable as high latency remains in that test ? Or is this in someway an issue with the router handling its power save (something like APSD etc) ? I also tried higher packet sized pings, 1024 and 2048 and still would not reduce latency. Only traffic away from the Mini’s, Mini to/from NAS, Mini running Speedtest etc seemed to reduce it.

2) Has anyone else here connected an AC capable Mac to an Asus router and can confirm Link Speed was 1 Gbit/s and or if they are seeing similar latency issues ?

Thanks

 

[Wildcatz_1]

Here are pics 3 & 4

 

[Wildcatz_1]

finally pic 5, sorry for having to separate out