Solved

Unmanageable Latency


Over the last couple weeks my Shadow latency has been spiking an hovering in the 700 m/s  range periodically.

 

This is not a local router issue as I tested both systems independently and my latency test on my base system is in the 50 m/s range. Once I speed test from Shadow there is a remarkable difference.

 

I primarily use the shadow for PC gaming and the effect is that I am playing in tar where my actions and results are disconnected in a very noticeable way and the screen will start to artifact with a smoky wavy digital noise that is only on the shadow.

 

My first thought was that global bandwidth may have spilled over into the connectivity but this problem has been very constant for about 3 weeks now. I cannot realistically play ANY first person shooter as the latency makes it a frustrating experience. I live in Los Angeles presumably close to a data center so Idon’t know what gives.

 

Any help in this arena would be appreciated.

icon

Best answer by Pettman 17 July 2020, 14:18

Have you tried lowering it to 15Mbps and enabling H265?

Also, do you have many people either streaming or downloading on your home network?

View original

26 replies

Badge

Being in the US, I guess you are still on Boost and have’t been given Ultra early?

 

Take a look a look at the speed tests here and us the test for your closest DC @Banzai:
https://help.shadow.tech/hc/en-gb/articles/360011196119-How-to-Test-Your-Connection-to-Shadow

Here is a capture of my BASE system numbers on the speed test.

 

 

And here is the result I get running the test out of Shadow.

Shadow Speed

As you can see, the difference is remakable. This is me testing on the LA data center from LA.

As for boost vs ultra. I signed up almost a year ago when there was not a tier system so whatever was default transition is what I have. Long time customer.

I will try to capture a screen with the digital noise and the latency in Shadow system management overlay (Alt-windows-o).

 

The digital artifacts made it seem like it was a stream encode issue but the delay in reaction make more sense with extreme latency.

I just checked my profile and I am at the Ultra level price tier.

 

What I have tried so far:

Reset router, unplug router (power and ethernet)

Nvidia driver updates on both systems

I disabled a ton of memory hogs in the windows start up.

I have started shadow with a clean base system startup.

I have lowered the bandwidth allowance to 25 MB and to some effect it has reduced the impact. It will have less noise in the signal but there will still be stutters in the video. Audio is typically fine in real time.

I flushed my ip through the command line.

 

 

Latency in shadow app:

Latency

Noise in the video

Noise

 

Badge

Have you tried lowering it to 15Mbps and enabling H265?

Also, do you have many people either streaming or downloading on your home network?

As peterman said and also I would try and run a constant ping on your local machine which is ping 8.8.8.8 -t while you are on your shadow to see if any of the latency is unusual

Have you tried lowering it to 15Mbps and enabling H265?

Also, do you have many people either streaming or downloading on your home network?

I have not lowered past 25 Mbps but I will try 15 Mbps and H265.

I did try H265 by itself initially but it didn’t seem to help at the time, but perhaps in conjunction with the lower overhead speed something might click. Thank you for the suggestion.

 

The only other access to my home network is for my Nest and Google Home and whatever process they run but no one else on the network is downloading or streaming content. I don’t even have my cell phone on the home network.

 

Again, thank you for the feedback.

 

 

Userlevel 4
Badge +3

@Banzai Just to clear something up...for the speed test results that you posted, the first one is from your local computer to the Shadow data center (which is in Santa Clara by the way, not LA), showing about 250Mbps with ~ 23ms of latency. The second one you did from your Shadow VM, is going to be fast like that (with virtually no latency), because the source and destination endpoints are in the same data center location (like doing a local speed test to your own router over Ethernet).

I don’t know what’s causing your spikes, but wanted to make sure this was understood.

Userlevel 4
Badge +3

One other thing to mention...moderators on this forum sometimes recommend fetching the public IP from your Shadow VM (browsing to a site like https://ipchicken.com/), and then using a tool like https://www.pingplotter.com/ to do a trace from your local computer to the Shadow public IP. Doing this isn’t “foolproof” because it’s only really testing ICMP traffic, which is a different protocol than what the streaming traffic uses. Still, it can sometimes detect general congestion along the path.

As peterman said and also I would try and run a constant ping on your local machine which is ping 8.8.8.8 -t while you are on your shadow to see if any of the latency is unusual

I ran a ping on my local machine when the shadow connection was acting up. Local system looks pretty constant. see below. thank you for the recommendation

 

@BanzaiJust to clear something up...for the speed test results that you posted, the first one is from your local computer to the Shadow data center (which is in Santa Clara by the way, not LA), showing about 250Mbps with ~ 23ms of latency. The second one you did from your Shadow VM, is going to be fast like that (with virtually no latency), because the source and destination endpoints are in the same data center location (like doing a local speed test to your own router over Ethernet).

I don’t know what’s causing your spikes, but wanted to make sure this was understood.

Thank you Jim. I did have that realization after I posted.

One other thing to mention...moderators on this forum sometimes recommend fetching the public IP from your Shadow VM (browsing to a site like https://ipchicken.com/), and then using a tool like https://www.pingplotter.com/ to do a trace from your local computer to the Shadow public IP. Doing this isn’t “foolproof” because it’s only really testing ICMP traffic, which is a different protocol than what the streaming traffic uses. Still, it can sometimes detect general congestion along the path.

Thank you Jim29er. 

 

I ran a trace to shadow.tech and here was the result. both times step 2 timed out, not sure what impact that has here but here are the hops.

 

Userlevel 4
Badge +3

@Banzai You performed that traceroute inside your Shadow VM; you need to do that on your local computer. The idea is to test the Internet path between your local computer and the SJ data center, because that’s the same path the stream takes when using your Shadow.

Those “timed out” hops just means there was a device at that point that didn’t return ICMP replies, which is normal and can be ignored. If one of them intermittently times out vs. returning replies, that can be an indication of congestion.

Lastly, doing the traceroute doesn’t provide an exhaustive test of the path...it just shows what the path is. You would need to use a tool like PingPlotter (or on Windows, a command line tool like pathping) to do more analysis. But as I mentioned before, these tools only use ICMP traffic, which is different that the streaming traffic that Shadow uses...that happens via a different protocol, which is more difficult to test.

@BanzaiYou performed that traceroute inside your Shadow VM; you need to do that on your local computer. The idea is to test the Internet path between your local computer and the SJ data center, because that’s the same path the stream takes when using your Shadow.

Those “timed out” hops just means there was a device at that point that didn’t return ICMP replies, which is normal and can be ignored. If one of them intermittently times out vs. returning replies, that can be an indication of congestion.

Lastly, doing the traceroute doesn’t provide an exhaustive test of the path...it just shows what the path is. You would need to use a tool like PingPlotter (or on Windows, a command line tool like pathping) to do more analysis. But as I mentioned before, these tools only use ICMP traffic, which is different that the streaming traffic that Shadow uses...that happens via a different protocol, which is more difficult to test.

I misunderstood. Here are the ping hops on my local system

 

Userlevel 4
Badge +3

@BanzaiIf you can, please install PingPlotter on your local computer, and then use it to do a trace to 170.249.94.90. Let it run for a good ten minutes, and then post a screenshot.

If you don’t want to install PingPlotter, you can run this from a command prompt on your local computer:

pathping 170.249.94.90

That will generate similar results, just not as pretty. :slight_smile:

@BanzaiIf you can, please install PingPlotter on your local computer, and then use it to do a trace to 170.249.94.90. Let it run for a good ten minutes, and then post a screenshot.

If you don’t want to install PingPlotter, you can run this from a command prompt on your local computer:

pathping 170.249.94.90

That will generate similar results, just not as pretty. :slight_smile:

I ran the ping path and here was the result. thank you for the guidance Jim29er.

 

 

Userlevel 4
Badge +1

Have you tried lowering it to 15Mbps and enabling H265?

Also, do you have many people either streaming or downloading on your home network?

he complains about latency and your recommendation is using H265? You do know, that this is even slower right? How can you recommend it?

Userlevel 4
Badge +3

I ran the ping path and here was the result. thank you for the guidance Jim29er.

 

Thank you @Banzai, from those results it doesn’t seem like network issues are to blame. Again, the ICMP is not an “exhaustive” test, but if there were UDP stream throttling/congestion, I’d think that’d be more apparent in your Shadow stats. Also as I understand, the suggestion to lower your bandwidth allocation and/or try H265 are efforts to improve latency dropouts, but being that yours are “intermittent” perhaps that doesn’t really apply.

One thing that I don’t think any of us have asked you, is what input method you’re using to play your game(s) - I’m assuming keyboard/mouse for the FPS vs. a controller, but I’m also wondering if a change in input method would change any behavior. Or perhaps another way of asking - do the spikes correlate at all to [heavy] mouse movement (or some other kind of action), or are they random?

Happy Friday!

I ran the ping path and here was the result. thank you for the guidance Jim29er.

 

 

One thing that I don’t think any of us have asked you, is what input method you’re using to play your game(s) - I’m assuming keyboard/mouse for the FPS vs. a controller, but I’m also wondering if a change in input method would change any behavior. Or perhaps another way of asking - do the spikes correlate at all to [heavy] mouse movement (or some other kind of action), or are they random?

Happy Friday.

 

I am using a mouse and keyboard combo but the issue is non related. The time window has coincided with peek internet traffic (8 pm- 11 pm) so I’m starting to think the backbone is getting slammed. Pandemic, am I right?

 

So I don’t know if this can be fixed, which would not be ideal as it would render streaming pc’s ineffective for twitch gaming. It’s disappointing because it only recently got to a place where the experience is absolutely unplayable but here we are.

 

It was a good run. I’ll see if tech support can do anything and if not just pause my account until things settle down.

 

Userlevel 3

@Banzai You might have already messed with this setting, but it’s worth mentioning. Have you tried setting the “Optimize Streaming” option to “Prefer speed”? I had the same issue as you when I had it set to “Prefer reliability”, with my latency spiking to like 1000ms. Ever since I’ve had it on “Prefer speed” my latency has been pretty consistent

Userlevel 6
Badge +5

I ran the ping path and here was the result. thank you for the guidance Jim29er.

 

 

One thing that I don’t think any of us have asked you, is what input method you’re using to play your game(s) - I’m assuming keyboard/mouse for the FPS vs. a controller, but I’m also wondering if a change in input method would change any behavior. Or perhaps another way of asking - do the spikes correlate at all to [heavy] mouse movement (or some other kind of action), or are they random?

Happy Friday.

 

I am using a mouse and keyboard combo but the issue is non related. The time window has coincided with peek internet traffic (8 pm- 11 pm) so I’m starting to think the backbone is getting slammed. Pandemic, am I right?

 

So I don’t know if this can be fixed, which would not be ideal as it would render streaming pc’s ineffective for twitch gaming. It’s disappointing because it only recently got to a place where the experience is absolutely unplayable but here we are.

 

It was a good run. I’ll see if tech support can do anything and if not just pause my account until things settle down.

 

If youre on the California data center, we have performed quite a few trace routes in the discords with a lot of users deporting packet loss and ping spikes. The conclusion was that the routes used by these users to the Shadow data center were very badly congested. You might be one of them. 

@Banzai You might have already messed with this setting, but it’s worth mentioning. Have you tried setting the “Optimize Streaming” option to “Prefer speed”? I had the same issue as you when I had it set to “Prefer reliability”, with my latency spiking to like 1000ms. Ever since I’ve had it on “Prefer speed” my latency has been pretty consistent

Mist - Fantastic suggestion. I had not gone back to the initial settings until you mentioned it. I flipped it to “Prefer speed” and ever since the experience has become manageable.

 

This appears to be revolved.

 

Thank you to the community to the feedback and response.

Userlevel 3

Mist - Fantastic suggestion. I had not gone back to the initial settings until you mentioned it. I flipped it to “Prefer speed” and ever since the experience has become manageable.

 

This appears to be revolved.

 

Thank you to the community to the feedback and response.

Sweet! Glad it worked for you. It’s kind of a weird fix, given that you’d expect “Prefer reliability” to be the option that would provide more consistent latency.

Reply