DLNA buffering on gigabit network

For help and support with Universal Media Server
Forum rules
Please make sure you follow the Problem Reporting Guidelines before posting if you want a reply
Post Reply
zeus0r
Posts: 2
Joined: Fri Feb 09, 2018 2:56 am

DLNA buffering on gigabit network

Post by zeus0r »

hey there!

following setup:

win7 pc with ultimate media server ---------------- unitymedia connectbox router (gigabit lan ports) ---------------- philips 55pos901 tv (everything connected via ethernet cable)

when playing high bitrate files (~70 Mbit/s) on my TV via DLNA it starts buffering after some time. (pretty randomly - sometimes after a few minutes, sometimes after an hour)

what could cause this? i mean 0,07 GBit/s shouldn't be a bandwith problem on a gigabit network, right? any idea what's the problem there?

thanks
Nadahar
Posts: 1990
Joined: Tue Jun 09, 2015 5:57 pm

Re: DLNA buffering on gigabit network

Post by Nadahar »

It shouldn't, but there are a number of things that could impact this. My first suspect would be the ethernet port on the TV. Do you know that it is gigabit? I took a quick look at the TV's specs, and they conveniently fails to mention this. In the user manual they recommend using Cat 5e, which would indicate that it is only a 100 Mbps port.

Your network might not actually run at gigabit speed either, depending on your cabling. Only Cat 6 cables should be used, and they should not run along power lines, hard bends etc. If packet loss is high (due to interference/crosstalk), the speed will be reduced.

It might also be, if you're transcoding, that your computer can't keep up. If you're not transcoding it could also be due to a very slow disk access, if for example the media files are placed at another device on the network.

Since you have problems around 70 Mbps, I suspect that something is running in 100 Mbps. 70 Mbps is typically around the maximum speed you can get out of a 100 Mbps connection.
Last edited by Nadahar on Thu Mar 21, 2019 5:44 am, edited 1 time in total.
zeus0r
Posts: 2
Joined: Fri Feb 09, 2018 2:56 am

Re: DLNA buffering on gigabit network

Post by zeus0r »

Nadahar wrote: Fri Feb 09, 2018 4:46 am It shouldn't, but there are a number of things that could impact this. My first suspect would be the ethernet port on the TV. Do you know that it is gigabit? I took a quick look at the TV's specs, and they conveniently fails to mention this. In the user manual they recommend using Cat 5e, which would indicate that it is only a 100 Mbps port.
wow, you got it on your first guess. i just installed some speedtest app and it actually peaked at exactly 60 Mbps. thanks for your help mate!
Nadahar
Posts: 1990
Joined: Tue Jun 09, 2015 5:57 pm

Re: DLNA buffering on gigabit network

Post by Nadahar »

That's good to hear. Sadly, manufacturers have a tendency to "forget" to mention the not-so-good specs, which also made my suspicious. Strangely, the Wifi NIC is much more capable than your wired NIC (that's a first for me). That means that by using a AC router, as long as your TV has a a strong signal, you should be able to get better bandwidth that way. They save money on the most insane things...
Leopardi
Posts: 11
Joined: Fri Jun 09, 2017 6:17 am

Re: DLNA buffering on gigabit network

Post by Leopardi »

Nadahar wrote: Fri Feb 09, 2018 4:46 am It shouldn't, but there are a number of things that could impact this. My first suspect would be the ethernet port on the TV. Do you know that it is gigabit? I took a quick look at the TV's specs, and they conveniently fails to mention this. In the user manual they recommend using Cat 5e, which would indicate that it is only a 100 Mbps port.

Your network might not actually run at gigabit speed either, depending on your cabling. Only Cat 6 cables should be used, and they should not run along power lines, hard bends etc. If packet loss is high (due to interference/crosstalk), the speed will be reduced.

It might also be, if you're transcoding, that your computer can't keep up. If you're not transcoding it could also be due to a very slow disk access, if for example the media files are placed at another device on the network.

Since you have problems around 70 Mbps, I suspect that something in running in 100 Mbps though. 70 Mbps is typically around the maximum speed you can get out of a 100 Mbps connection.
Why do you say that? Cat 5e is rated for gigabit in up to 100 meter distance...

I bought a new cable router which has gigabit ports so I could up the UMS settings from 100mbit to 1000mbit, are you saying that using the 5e cable would not be enough? Tech specs say that it is enough.

But where has the speed test disappeared from logs? How do I check speed?
Nadahar
Posts: 1990
Joined: Tue Jun 09, 2015 5:57 pm

Re: DLNA buffering on gigabit network

Post by Nadahar »

Theoretically Cat5e can handle 1 Gbps and Cat6 can handle 10 Gbps, so you're correct that Cat5e should do it. Personally I have had different experience, and I exclusively use Cat6 and has replaced (most of) my fixed Cat5e cabling in my house. The think is that these standards basically decide how resistant the cable is to noise/interference and crosstalk. While crosstalk shouldn't be a problem, since that only depends on the speed itself, noise/interference can be. I've thus noticed that in the real world Cat5e often won't handle 1 Gbps. Network equipment negotiate speed based on their capabilites and how "clean" the signal is. That means that even if all your equipment and cables are up to spec, it will actually run at a lower speed if the signal isn't up to pair. There are many requirements for how network cables should run to achieve their specification, especially when it comes to magnetic interference. They should for example NOT run parallel to mains AC cables, since the polarity switching (of the AC) creates magnetic interference. How many people can achieve this in the real world? Likewise, there could be other sources of noise that the cables run by etc. Thus, I've found that the "extra protection" Cat6 gives come in handy also for 1 Gbps.

In addition to the negotiate speed, if the interference varies a lot the negotiated speed might be "too high" for the worst case scenarios. That will lead to packet loss which means that packets must be retransmitted, which effectively lowers the speed below the negotiated speed.

That said, I don't think the cabling should be the prime suspect here. I think the Ethernet port on the renderer is the most likely bottleneck. Since gigabit ports must be must more "precise" they probably require better materials or a more elaborate design. I don't know the details, but I know that they are more expensive for the manufacturer. They often choose to "save costs" on things like that which is hard for the consumer to verify/evalutae, and being vague about the capabilities can often be a bad sign. When they choose the high-end/expensive option, they usually aren't shy about mentioning it.

Remember that for a connection to achieve 1 Gbps, both ends and the cabling must allow it. It doesn't matter if you have a 10 Gbps router in one end if the other end runs 100 Mbps, the connection will still be 100 Mbps.

Regarding the "speed tests" in the log, I think they only appear if your have "automatic bandwidth" enabled. That said, you should take the results with a grain of salt, as the ability to measure this from UMS' POV is quite limited. Think of it as an approximation, and one that is only valid at the exact time which it is measured. You should be able to use it to determine if the capability is vastly different than expected though.
Leopardi
Posts: 11
Joined: Fri Jun 09, 2017 6:17 am

Re: DLNA buffering on gigabit network

Post by Leopardi »

Nadahar wrote: Thu Mar 21, 2019 6:09 am Theoretically Cat5e can handle 1 Gbps and Cat6 can handle 10 Gbps, so you're correct that Cat5e should do it. Personally I have had different experience, and I exclusively use Cat6 and has replaced (most of) my fixed Cat5e cabling in my house. The think is that these standards basically decide how resistant the cable is to noise/interference and crosstalk. While crosstalk shouldn't be a problem, since that only depends on the speed itself, noise/interference can be. I've thus noticed that in the real world Cat5e often won't handle 1 Gbps. Network equipment negotiate speed based on their capabilites and how "clean" the signal is. That means that even if all your equipment and cables are up to spec, it will actually run at a lower speed if the signal isn't up to pair. There are many requirements for how network cables should run to achieve their specification, especially when it comes to magnetic interference. They should for example NOT run parallel to mains AC cables, since the polarity switching (of the AC) creates magnetic interference. How many people can achieve this in the real world? Likewise, there could be other sources of noise that the cables run by etc. Thus, I've found that the "extra protection" Cat6 gives come in handy also for 1 Gbps.

In addition to the negotiate speed, if the interference varies a lot the negotiated speed might be "too high" for the worst case scenarios. That will lead to packet loss which means that packets must be retransmitted, which effectively lowers the speed below the negotiated speed.

That said, I don't think the cabling should be the prime suspect here. I think the Ethernet port on the renderer is the most likely bottleneck. Since gigabit ports must be must more "precise" they probably require better materials or a more elaborate design. I don't know the details, but I know that they are more expensive for the manufacturer. They often choose to "save costs" on things like that which is hard for the consumer to verify/evalutae, and being vague about the capabilities can often be a bad sign. When they choose the high-end/expensive option, they usually aren't shy about mentioning it.

Remember that for a connection to achieve 1 Gbps, both ends and the cabling must allow it. It doesn't matter if you have a 10 Gbps router in one end if the other end runs 100 Mbps, the connection will still be 100 Mbps.

Regarding the "speed tests" in the log, I think they only appear if your have "automatic bandwidth" enabled. That said, you should take the results with a grain of salt, as the ability to measure this from UMS' POV is quite limited. Think of it as an approximation, and one that is only valid at the exact time which it is measured. You should be able to use it to determine if the capability is vastly different than expected though.
I connected it with the cat5e to the new gigabit router and it reports about 282 mbit/s, from 65mbit/s. I guess that would be enough for using the gigabit setting, streaming 1080p or even 2160p files, without having to get a new cable?

I remember using the lossless manual settings with the 100mbit and it seemed to work effectively.

E: Interestingly, just connected to PS3 it reports PS3 at 282mbit and windows media player at 700mbit?
Nadahar
Posts: 1990
Joined: Tue Jun 09, 2015 5:57 pm

Re: DLNA buffering on gigabit network

Post by Nadahar »

What is reporting 282 mbps? It's a strange number to get from a gigabit network, although 65 mbps would be reasonable for a 100 mbps network. Are you referring to UMS' "speed check"? If so, it's not so strange as it's not suitable to measure such high speeds, as it uses very small packets when measuring, meaning the accuracy falls as the speed increases.

I assume that WMP is running on the same computer as UMS, which would explain why the speed is higher. Because these speeds are outside the range for the UMS' "speed check", what you're actually seeing here is a difference in latency. There is a bit of "overhead" in sending a packet, and for the other party to reply back. The time of this overhead becomes relatively large when the transfer of the packet takes very little time. Sending a packet "internally" on the same computer skips a lot of steps, and thus the overhead is much less - which would explain why it "measures" higher speed.
Leopardi
Posts: 11
Joined: Fri Jun 09, 2017 6:17 am

Re: DLNA buffering on gigabit network

Post by Leopardi »

Nadahar wrote: Fri Mar 22, 2019 5:35 am What is reporting 282 mbps? It's a strange number to get from a gigabit network, although 65 mbps would be reasonable for a 100 mbps network. Are you referring to UMS' "speed check"? If so, it's not so strange as it's not suitable to measure such high speeds, as it uses very small packets when measuring, meaning the accuracy falls as the speed increases.

I assume that WMP is running on the same computer as UMS, which would explain why the speed is higher. Because these speeds are outside the range for the UMS' "speed check", what you're actually seeing here is a difference in latency. There is a bit of "overhead" in sending a packet, and for the other party to reply back. The time of this overhead becomes relatively large when the transfer of the packet takes very little time. Sending a packet "internally" on the same computer skips a lot of steps, and thus the overhead is much less - which would explain why it "measures" higher speed.
PS3 in UMS speed check reports 282mbit/s. I'll just assume it's doing near max gigabit then.

I guess the automatic gigabit option runs the lossless transcode setting, so it's no use to start tinkering with it?
Nadahar
Posts: 1990
Joined: Tue Jun 09, 2015 5:57 pm

Re: DLNA buffering on gigabit network

Post by Nadahar »

Leopardi wrote: Fri Mar 22, 2019 9:35 am PS3 in UMS speed check reports 282mbit/s. I'll just assume it's doing near max gigabit then.
Yes, that sounds reasonable.
Leopardi wrote: Fri Mar 22, 2019 9:35 am I guess the automatic gigabit option runs the lossless transcode setting, so it's no use to start tinkering with it?
I'm not quite sure what those options does for the transcoding, somebody else will have to answer that.
Post Reply