Ping latency

[david 929]

Another data point.  I tested a Pixel 4 XL (android 10) and it gets about 2-3 ms to the wireless access point, best case scenario. 

For anyone else testing, if the router is busy with a lot of other devices talking to it, or you have a lot of apps running on the phone at the same time, that can affect the speed too.

[EskeRahn 5,454]

2 hours ago, david said:

Another data point.  I tested a Pixel 4 XL (android 10) and it gets about 2-3 ms to the wireless access point, best case scenario. 

For anyone else testing, if the router is busy with a lot of other devices talking to it, or you have a lot of apps running on the phone at the same time, that can affect the speed too.

Just a thought. If anyone got a usb-c adapter with a lan-plug, it could be interesting to see the times here with a wired connection..

[_DW_ 628]

43 minutes ago, EskeRahn said:

Just a thought. If anyone got a usb-c adapter with a lan-plug, it could be interesting to see the times here with a wired connection..

Could try plugging it into my laptop dock later 😄

[david 929]

7 hours ago, EskeRahn said:

Just a thought. If anyone got a usb-c adapter with a lan-plug, it could be interesting to see the times here with a wired connection..

I did use USB tethering to create a network with my laptop and that showed low pings.  But another form of that would be a good test.

[SirBaconIII 66]

Here's another data point. The second time pinging the gateway is consistantly low, in the 2ms range for me. I could be misreading it though. Also, this is initial testing on my work network. I'd be more confident in my home network where I know how traffic is routed.

Edited January 31, 2020 by SirBaconIII
Swap picture

[david 929]

1 hour ago, SirBaconIII said:

Here's another data point. The second time pinging the gateway is consistantly low, in the 2ms range for me. I could be misreading it though. Also, this is initial testing on my work network. I'd be more confident in my home network where I know how traffic is routed.

Those are interesting results, but traceroute does weird things with calculating the latency based on TTL, I think, vs actual pings.  What do you get if you ping those DNS servers instead of using traceroute?

You mentioned pinging the gateway, but I don't see that test.

UPDATE:  I just realized it is tracepath,  not traceroute.   I don't have that, since I'm on Windows.  Is it showing the delta between each hop?  It must be, since later hops should take longer to get to.  If so, then that does make sense.  The first hop (to the gateway) is 20+ ms, which is consistent with what we've been seeing.  And then the other hops are deltas on top of that, but since they are between each hop, it isn't including the overhead of previous hops and not including the latency issue the Pro1 has.

That second entry for the gateway is probably just the gateway talking to itself on different interfaces (although you'd think that would have a different IP address) or something.  It is inside the gateway, so is just a delta between them, I'm guessing.

Edited January 31, 2020 by david

[sdx 45]

There is something funky with the Wifi driver currently. This seems to relate mainly to roaming.

Usually, this sounds like you're hitting buffered data that's sent when a timer expires: https://en.wikipedia.org/wiki/Nagle's_algorithm

Or you're just on 2,4GHz in a noisy environment. 🙂

I did however try myself with HE network tools. It gives me 1.23ms katency with 0.09ms jitter. This was measured over wifi on 5GHz only. More info about the connection:

Performance

Signal Strength: -54 dBm Signal Quality: 39 dB Connection Speed: 867 Mbps Channel Width: 80 MHz

Capabilities

802.11ac (5GHz) Spatial Stream: 2

[SchattengestaIt 559]

@david I don't have the Pro1 yet, so I can't compare right now. I'm just trying to help providing data.

If you know any website with 100+ elements, you could try if it makes a difference. Also, does ping work that way? Will all elements load one after each other or aren't the requests sent all at the same time and - given the sufficient bandwidth - received at the same time (+ping =20ms)? That would mean that 20ms more wouldn't make any difference since it applies only once. I was surfing in sattelite internet for a time and had a ping of 800-1500ms. You notice that with websites but you don't sit there a minute waiting (40-70 elements).

Same thing in live games. 20ms more isn't a too big problem although pro gamers will discuss about that. But on mobile games ping shouldn't be that important (at least in the range we are talking about).

The problem with 5G will be that the small wavelength gets reflected and blocked much more easily by objects. That can block your signal entirely or reduce the bandwidth massively. I don't know how it affects ping, but I guess it will too. I don't see a future of 5G without 4G available when the signal drops.

 

Ans lastly, a slight problem can of course still be solved and you're absolutely right that it could help solving other problems as well. Good thinking there with the general wifi issues. I was just trying to understand what real life effects such a ping increase could have.

[david 929]

9 hours ago, SchattengestaIt said:

If you know any website with 100+ elements, you could try if it makes a difference.

100+ is the average number now for popular websites, unfortunately.

9 hours ago, SchattengestaIt said:

Also, does ping work that way? Will all elements load one after each other or aren't the requests sent all at the same time and - given the sufficient bandwidth - received at the same time (+ping =20ms)? That would mean that 20ms more wouldn't make any difference since it applies only once. I was surfing in sattelite internet for a time and had a ping of 800-1500ms. You notice that with websites but you don't sit there a minute waiting (40-70 elements).

That is a good point, but web browsers won't make 100+ connections concurrently. They may make a handful of connections concurrently though, which will reduce the issue to the degree that there is parallelism.  

9 hours ago, SchattengestaIt said:

The problem with 5G will be that the small wavelength gets reflected and blocked much more easily by objects. That can block your signal entirely or reduce the bandwidth massively. I don't know how it affects ping, but I guess it will too. I don't see a future of 5G without 4G available when the signal drops.

I am talking about 5G with a good signal.  Low latency is a key selling point of 5G.  It has been designed with this in mind.  The higher frequency 5G will not travel as far.  There will be lower frequency 5G as well.  Eventually, 5G will take over 4G frequencies.  6G will take over 5G, etc.  In other words, 5G is a method of communication.  It isn't tied to certain frequencies.  It just happens that the new frequencies that have been opened up were opened up when 5G came into being.  But in the USA, there are new low frequencies too, licensed by T-Mobile, in the 600 MHz range.  Of course, with greater distance allowed by the lower frequencies, latency will be higher.

What I am saying is that *if* the latency is also present on the cellular side of things in the Pro1, then that is another reason to look into the issue.  It might not be present on the cellular side at all.  It doesn't appear to be present on the wired side (USB port).

9 hours ago, SchattengestaIt said:

Ans lastly, a slight problem can of course still be solved and you're absolutely right that it could help solving other problems as well. Good thinking there with the general wifi issues. I was just trying to understand what real life effects such a ping increase could have.

Minimal effects, but still worth fixing, in my view.  We have a powerful device.  We might as well have it load web pages as fast as it can and achieve its full potential.

There *might* be some sort of benchmarks where they test page load time and that could make the Pro1 appear less desirable compared to other phones, as a result of this issue.  I haven't looked into that.  It is just something I thought of.

[david 929]

10 hours ago, sdx said:

There is something funky with the Wifi driver currently. This seems to relate mainly to roaming.

Can you explain this in more detail?  What type of roaming are you referring to?

10 hours ago, sdx said:

Usually, this sounds like you're hitting buffered data that's sent when a timer expires: https://en.wikipedia.org/wiki/Nagle's_algorithm

Yeah, some sort of timer could be involved.

10 hours ago, sdx said:

Or you're just on 2,4GHz in a noisy environment. 🙂

This is a consistent issue though, with the latency around 20 ms.  With a congested environment, the values would jump around more.

10 hours ago, sdx said:

I did however try myself with HE network tools. It gives me 1.23ms katency with 0.09ms jitter. This was measured over wifi on 5GHz only. More info about the connection:

Performance

Signal Strength: -54 dBm Signal Quality: 39 dB Connection Speed: 867 Mbps Channel Width: 80 MHz

Capabilities

802.11ac (5GHz) Spatial Stream: 2

I downloaded the app to the Pro1, but I can't find which tool to run that gives a jitter value.  The Ping function in the HE network tools I downloaded runs 3 pings very fast (no delay, which is what other ping tools usually do) and gives an average, std. dev (maybe that is what you are referring to as jitter?), last, best and worst values.  For some reason, it does normally show 2-3 ms for me, which is much better than the normal value I get from other ping tools.

So the question is, what is it doing that the other tools are not doing?  It isn't delaying between pings, so maybe that has some effect on this behavior.

I did try this:

ping -i .2 192.168.1.3 

That ran the pings 200 ms apart instead of 1 second, but I still got 20 ms as a frequent latency.

[sdx 45]

7 minutes ago, david said:

Can you explain this in more detail?  What type of roaming are you referring to?

Yeah, some sort of timer could be involved.

This is a consistent issue though, with the latency around 20 ms.  With a congested environment, the values would jump around more.

I downloaded the app to the Pro1, but I can't find which tool to run that gives a jitter value.  The Ping function in the HE network tools I downloaded runs 3 pings very fast (no delay, which is what other ping tools usually do) and gives an average, std. dev (maybe that is what you are referring to as jitter?), last, best and worst values.  For some reason, it does normally show 2-3 ms for me, which is much better than the normal value I get from other ping tools.

So the question is, what is it doing that the other tools are not doing?  It isn't delaying between pings, so maybe that has some effect on this behavior.

I did try this:

ping -i .2 192.168.1.3 

That ran the pings 200 ms apart instead of 1 second, but I still got 20 ms as a frequent latency.

Roaming between AP's. I have three AP's in my apartment, controlled by one wireless controller. They're all Cisco 2700 series AP's.

Yes, jitter is the std deviation as seen in HE network tools. So given that HE network tools maybe sends three ICMP echo requests very fast, maybe that's enough to flush the buffer? In that case, this is more of a cosmetic problem in most cases, unless you want to infrequently transfer low amounts of data with low latency.

[DieBruine 397]

I've tried several pings using Fring. Only the first ping is 20ms. Out of 15 tries 9 times the first ping, never the second.

5GHz connection, tried 2 different brands, one of whick Unifi. Didn't seem to make any difference.

[david 929]

12 hours ago, sdx said:

Roaming between AP's. I have three AP's in my apartment, controlled by one wireless controller. They're all Cisco 2700 series AP's.

Yes, jitter is the std deviation as seen in HE network tools. So given that HE network tools maybe sends three ICMP echo requests very fast, maybe that's enough to flush the buffer? In that case, this is more of a cosmetic problem in most cases, unless you want to infrequently transfer low amounts of data with low latency.

Yeah, I guess that is possible.  Not a very big buffer then.  I just tried pinging with more data instead of the default amount and that didn't cause the latency to decrease.  It would have to be based on number of requests, rather than size of the data, to work like you theorized, I think.

ping -s 10000 192.168.1.3

That's what I ran, in case anyone else wants to try.  That sends 10000 bytes instead of the default 56 bytes (not including the IP header).

Edited February 2, 2020 by david

[david 929]

10 hours ago, DieBruine said:

I've tried several pings using Fring. Only the first ping is 20ms. Out of 15 tries 9 times the first ping, never the second.

5GHz connection, tried 2 different brands, one of whick Unifi. Didn't seem to make any difference.

Can you send the link to the app you used?  I wasn't able to find it.

[david 929]

I tried this tool:  https://play.google.com/store/apps/details?id=com.julian.fullping&hl=en_US

In the settings, you can set the delay between pings down to as little as 0ms.  Setting 1 ms shows that most pings finish in about 2 ms.  That would make sense with what we have seen with some of these other tools.  What doesn't make sense to me is if I set it to values higher than 2 ms, it still allows the pings to finish, most of the time, in about 2 ms.  For instance, I can set it to as high as about between 100 and 110 ms and it still has this effect.  How can a delay 50 times longer than the time it takes for the packet to go to the endpoint and back make the process take less time?

The more I think about this, it isn't the delay between ping responses.  The setting is telling it to send the packet every N ms, not to wait for the response.  So in that case, if there is some sort of counter that is watching how many requests have been made, and then releasing them, the first X requests would be held up, but then after that point, they would be sent out, and *possibly*, if this counter is watching a sliding window, then that is what keeps it sending them out quickly instead of waiting for another batch to build up.  Because if it had to wait for X requests before it would send, then it would only be a few near the end of the batch of requests that would be lower in response time.  

Either way, this does show that the delay between sending requests does affect the behavior.

Now, in a real world situation, on a web page, for instance, how would this affect things?  Let's say we have as 5 threads in a browser sending out requests for different objects, concurrently.  So 5 are launched at the same exact time at the start.  If 5 is equal to or more than however many it takes to trigger the system to release the batch, then they go out instantly.  They will each come back after a different amount of time.  This will cause the following requests to be spaced out more, which could lead us back to the delay issue.  Because, unlike pinging an access point that is only 1-5 ms away, objects on the internet might be 30 to 300 or more ms away. 

The more threads there are, the greater chance we can get to X faster each time and cause the batch to be released sooner.

All of this is assuming it is requests per unit of time that is the deciding factor.  From my testing, the amount of data doesn't seem to be the deciding factor.  I can't think of what else would fit what we are seeing, but feel free to throw ideas out there if anyone has any other ideas.

[DieBruine 397]

1 hour ago, david said:

Can you send the link to the app you used?  I wasn't able to find it.

My bad, typo 🙄

https://play.google.com/store/apps/details?id=com.overlook.android.fing

[SchattengestaIt 559]

@david (sorry that I can't multi-quote on the phone)

True, you are probably right about the 100 elements on a website. A lot is handled with scripts, but I don't know how multiple pictures are handled since most of them come from the same server. I am looking forward into using uMatrix on the phone in order to block most scripts anyways, but that's not a topic for this thread.

You are also right about the 5G with different frequencies (and therefore latencies/ping). However, currently both "systems" are built parallel of each other and I also see 2G and 3G still working (while 3G is removed mostly in Germany). Including all frequencies into one transmitter mast would be great and would allow a much better user experience. But as phones get better, that will become less of a problem.

By the way, in Germany I haven't heard of any MHz-frequencies, as far as I know, 3-4GHz and 12+ GHz were sold. So here it's really about range.

I agree on your other points, it's worth looking into it and making it better (as it is with everything in life)

 

[david 929]

8 hours ago, SchattengestaIt said:

You are also right about the 5G with different frequencies (and therefore latencies/ping). However, currently both "systems" are built parallel of each other and I also see 2G and 3G still working (while 3G is removed mostly in Germany). Including all frequencies into one transmitter mast would be great and would allow a much better user experience. But as phones get better, that will become less of a problem.

Some providers are phasing out support for 2G-3G in the USA.  They are cannibalizing the old protocol frequencies to make room for serving more people on the newer protocols. 

8 hours ago, SchattengestaIt said:

By the way, in Germany I haven't heard of any MHz-frequencies, as far as I know, 3-4GHz and 12+ GHz were sold. So here it's really about range.

That's too bad.  I'm guessing that will reduce 5G adoption.  However, in reality, even in the USA, I think it will take a while for lower frequencies to be used by all providers for 5G use.

I did find this:  https://www.everythingrf.com/News/details/8554-Germany-Makes-700-MHz-Frequency-Band-Available-for-5G-Rollouts

Quote

The 700 MHz band, previously had been used for terrestrial television services, however, those services have now been migrated to different frequencies. According to BNetzA President, Jochen Homann, the release of the 700 MHz spectrum band would help Germany's mobile network operators offer a more comprehensive service. The additional frequencies are now available nationwide for mobile network operators to improve their coverage, especially in rural areas, according to him.

and

Quote

EU member states are obliged to free up the 700 MHz band spectrum for 5G usage by the end of June 2020. A deadline of 30 June 2020 (or up to 2 years later with adequate justification) has been set for the reassignment of the 700 MHz frequency band (694–790 MHz) to wireless broadband services in Europe. According to a report by Analysys Mason, originally, the 700 MHz band was envisaged as being suitable to provide additional 4G capacity and was seen as being particularly useful for this purpose in rural areas or towards cell edges. However, more recently, policy makers have been assuming that the 700 MHz band will provide the wide area coverage layer for 5G new radio (NR). The European Commission, for example, has a target of ‘5G everywhere’ by 2025, which, to be achievable, would need to make heavy use of low frequency spectrum (and hence the 700 MHz band).

So it looks like you didn't have the 700 MHz bad available in Germany before, for cellular, whereas in the USA, that is one of the main bands used for covering customers at longer distances.  But you should be getting it, starting in the second half of 2020, unless all the current users of 700 MHz get a 2 year delay. :-)  What they said at the end also makes it look like there will be more lower frequencies opened up so that they can meet the goal to provide 5G everywhere.

[SchattengestaIt 559]

@david I'd like to explain Germany when it comes to politics and european rules.

If we are lucky, they will indeed be implemented in time. In this case, since the frequencies weren't used before, it's easy to do.

But then comes the usage. In Germany, frequencies are sold when new technology is introduced (I guess that's the same everywhere). The last auction had the previously named frequencies, 3-4GHz and I think 6-12GHz. Higher frequencies could be sold later, I haven't read anything about a MHz range (although that doesn't mean that it couldn't come).

The EU goal meanwhile is to provide 5G everywhere, which would be easiest with low frequencies, but the providers can earn more money from the politics if they don't tell them and just place more transmitters. Many billions were just paid in order to complete that task. That's at least my theory why we still have no connection in many rural areas while it would be relatively easy to do.

So a goal is a goal but I'm pretty sure that it will not be achieved.

 

I do still thank you for your research and I learned a lot in this discussion.

Edited February 3, 2020 by SchattengestaIt

[archseraphim 50]

I'm finding poor local wifi performance as well.  Pinging the phone from a machine on my local network gives me an average ping of 82ms(!). Phone is on 5GHz wifi, other machine is wired.

The performance of the Steam Link app, which is hugely sensitive to network conditions, is also abysmal.

[david 929]

2 hours ago, archseraphim said:

I'm finding poor local wifi performance as well.  Pinging the phone from a machine on my local network gives me an average ping of 82ms(!). Phone is on 5GHz wifi, other machine is wired.

Yeah, we've been talking about pinging out from the phone, but pinging from another machine to the phone also has extra latency vs pinging other phones.

2 hours ago, archseraphim said:

The performance of the Steam Link app, which is hugely sensitive to network conditions, is also abysmal.

With this setup, do the games run right on your PC, or is the PC connecting to the internet for accessing the games?

Am I correct that it is mirroring the screen of the PC to your phone screen, so it needs to send video to the phone and you need both high bandwidth and low latency to keep up with the state of the game?

Is he wireless access point running 802.11n or 802.11ac?

I might have an idea for a test you could do that might provide a working solution.

[archseraphim 50]

1 hour ago, david said:

With this setup, do the games run right on your PC, or is the PC connecting to the internet for accessing the games?

Am I correct that it is mirroring the screen of the PC to your phone screen, so it needs to send video to the phone and you need both high bandwidth and low latency to keep up with the state of the game?

Is he wireless access point running 802.11n or 802.11ac?

I might have an idea for a test you could do that might provide a working solution.

Yes, PC to Phone, all using the local network. Traffic over the internet is not involved.

The network is 802.11ac. Running the Steam Link app on an Android TV device connected to the same network is a near-flawless experience.

[david 929]

12 minutes ago, archseraphim said:

Yes, PC to Phone, all using the local network. Traffic over the internet is not involved.

The network is 802.11ac. Running the Steam Link app on an Android TV device connected to the same network is a near-flawless experience.

I would try connecting the phone to your PC, using a USB 3 port, and then in the USB Preferences screen, in the phone, turn on USB tethering.  You shouldn't have to enable USB debugging first, but if it doesn't work for some reason, you can try that.  Your PC (assuming windows) will pop up asking if you want the network to be a Home, Work or Public network. You can pick Home, since nothing can spy on that network anyway, since it is only between the phone and the PC, over the USB cable.

You may need to disable your wifi on the PC (if is running wifi) or unplug the ethernet cable (if it is connected via ethernet) in order to get things to work.

When it is all set up, if the Steam Link app normally auto-discovers the PC app, then it should auto-discover it and work. If it doesn't auto-discover it and has a place for you to enter the IP address of the PC, then you can follow the directions in this post to get to the IP address of the PC, in this tethered network:

This may not be a permanent solution for you, because you might not want to be tethered to your PC when you are playing the games, but at least it will tell you if it is the network that is causing the problems or something else with the phone.  The network you'll get between the two will be plenty fast and have very low latency (1-5 ms pings).

[archseraphim 50]

13 hours ago, david said:

I would try connecting the phone to your PC, using a USB 3 port, and then in the USB Preferences screen, in the phone, turn on USB tethering.  You shouldn't have to enable USB debugging first, but if it doesn't work for some reason, you can try that.  Your PC (assuming windows) will pop up asking if you want the network to be a Home, Work or Public network. You can pick Home, since nothing can spy on that network anyway, since it is only between the phone and the PC, over the USB cable.

This may not be a permanent solution for you, because you might not want to be tethered to your PC when you are playing the games, but at least it will tell you if it is the network that is causing the problems or something else with the phone.  The network you'll get between the two will be plenty fast and have very low latency (1-5 ms pings).

I'll give that a spin, though you're correct about it not being a permanent solution. The entire point of Steam Link is that I can play my games while away from my computer, after all.

I wouldn't consider 1-5ms "very low latency" on a local network though. Acceptable on wifi, sure, but not very low.

[david 929]

2 hours ago, archseraphim said:

I wouldn't consider 1-5ms "very low latency" on a local network though. Acceptable on wifi, sure, but not very low.

You are right. I think I meant compared to how it is normally.