Buffering issue when connected via switch

Worth a try . . . I had similar issues and by process of elimination determined it was my Airport Extreme (that was being used as my wifi source).  I reset the Airport to factory settings (you can read online how to do this) and that fixed it.  I cannot explain what was wrong with it, but that was it.

I would try that with both of your Airports; it can't hurt.

Worth checking weather there is a mismatch on the supported port speeds. This sounds very much like the buffer memory in the switch itself is full. This happens for several reasons, none of which are that easy to diagnose on a regular SOHO device.

A switch, router (or combination of the two) has a buffer memory to facilitate flow control. When this gets full, the transfer rate to one or more hosts will suffer and network performance will fall off a cliff. The reasons why the buffer fills up vary, with the only common point being that data is entering the unit faster than it can be sent out. Common reasons:

• Basic network congestion. Devices attached is streaming this and downloading that and the unit is maxed out.
• Different supported transfer speeds. Assuming all connected devices use auto negotiated duplexing and speed assignment (which is nearly always the case with consumer units) the switch may have to facilitate transfer between two endpoints of very different capability. For example, a NAS that supports gigabit Ethernet and another device that doesn't. This effectively causes flow control to occur which has an impact on the memory buffer in the switch.

Most low cost SOHO devices, even if gigabit supporting on all ports, only support this as an aggregate speed across everything. Still, streaming WAV at 24/192KHz doesn't actually take much bandwidth (nowhere near gigiabit).

With combined wifi access point/switches/routers the problem is exacerbated a bit because normally, to cut costs on cheaper units, the wifi connections support very high bandwidth close to gigabit Ethernet but the physical ports may be limited to 100BaseT. So in a home network, the throughput is going to be affected by a combination of the slowest link in the chain of devices the traffic must pass through combined with how well the memory buffer on each the switches in that route is holding up.

I imagine it is pretty hard to see what is going on inside an AirPort box or most bog standard consumer devices. Suffice to say that if there are a house full if devices (wired or otherwise), problems can arise quickly. Restarting a device dumps its buffer memory to zero. But more interestingly, restarting a device causes other network devices to dump memory that was enroute to the restarted device since those queued packets will time out.

thanks all, it's trial and error mode right now,

@Bart: i have indeed reset both AEs;

is there a software i can install on my PC which when the incident reoccurs can trace the route of data between the NAS and the NDX and detect whether -for whatever reason- the traffic goes back to the router (it shouldn't!) or indeed stays local (NAS -> SWITCH -> NDX) but perhaps then it is a buffer_memory issue as FEELING_ZEN writes? I have tried the simple tracert command in command promt but it does not tell me anything intelligent as i only have one router in this network and also downloaded a PRTG network utility which again does not seem to give me this information.

DrPo posted:

thanks all, it's trial and error mode right now,

@Bart: i have indeed reset both AEs;

is there a software i can install on my PC which when the incident reoccurs can trace the route of data between the NAS and the NDX and detect whether -for whatever reason- the traffic goes back to the router (it shouldn't!) or indeed stays local (NAS -> SWITCH -> NDX) but perhaps then it is a buffer_memory issue as FEELING_ZEN writes? I have tried the simple tracert command in command promt but it does not tell me anything intelligent as i only have one router in this network and also downloaded a PRTG network utility which again does not seem to give me this information.

You should be able to play music with just your streamer, NAS, and an Airport Express connected to a switch and everything else turned off. This will tell you if your router is doing something odd that's causing your problem. Not a very scientific approach, I know, but maybe worth a try.

I'm confused about the connections you are making. Are you able to explain again what is connected to what and whether they are on the same subnet? The need for the second AE is confusing to me. Special attention to any devices like the NAS which may have more than one network connection (wired and wifi) and devices that act as routers and have their own IP address.

A diagram of the logical connections (wired or otherwise) might help. Buffer memory on a network device still seems most likely to me but the setup seems a little convoluted. That could be further aggravating the issue or indeed part of a different root cause entirely.

DrPo, since you tried the replacement patch leads I suggested have you had a recurrence?

Simon:

DrPo, since you tried the replacement patch leads I suggested have you had a recurrence?also tried resetting both AEs, the issue still occurs.

I think I effectively did by replacing the cable itself - still have issues. Also reset both AEs and still have issues. When back from work this evening i will set fixed IP addresses to both switch and NDX and re-monitor.

feeling_zen posted:

I'm confused about the connections you are making. Are you able to explain again what is connected to what and whether they are on the same subnet? The need for the second AE is confusing to me. Special attention to any devices like the NAS which may have more than one network connection (wired and wifi) and devices that act as routers and have their own IP address.

A diagram of the logical connections (wired or otherwise) might help. Buffer memory on a network device still seems most likely to me but the setup seems a little convoluted. That could be further aggravating the issue or indeed part of a different root cause entirely.

 

yes, let me explain:

the reason for the 2 AEs is that i cannot afford a wired connection in the room (you can guess the reasons). So the set up is as follows

COMMENT: I don't activate the wireless of the vdsl modem as i had significant dropout problems with it for weeks, the AE solution works 2 years now without any issue)

ChrisSU posted: again does not seem to give me this information.

You should be able to play music with just your streamer, NAS, and an Airport Express connected to a switch and everything else turned off. This will tell you if your router is doing something odd that's causing your problem. Not a very scientific approach, I know, but maybe worth a try.

Chris, i was thinking along the same lines last night but could not make this test, i will retry tonight (btw i have posted the set up of my connections above )

DrPo posted:

When back from work this evening i will set fixed IP addresses to both switch and NDX and re-monitor.

Hi Mike, not sure if it's wise - but it's an easy thing to try, getting a different NAS proved to be more complex but it is also something i will eventually do (good stuff to play with during Xmas days off). 

DrPo - next time it happens - power down the AE. The streaming should continue - although the iPad wont display anything.

If the stuttering still happens then we can focus on the Streamer / Switch / NAS and wiring

Given your diagram - my suspicion is the NAS. We spoke on the other thread about you trying a different UPnP media server perhaps on a PC or Mac connected to the switch - I would like you to try and do that.

BTW I wouldn't recommend bypassing the DHCP - as Mike says you are inviting other challenges and really should not be relevant.

Simon

OK, will work on a NAS alternative and will not play with fixed IPs,

btw Simon, when you say " power down the AE" I presume you refer to AE(1) (which is the "active" one), right? 

Yes it can be either AE(1) and/or AE(2) in your diagram

Simon

Hi Simon, I could not "recreate" the issue last night (these are the most frustrating problems in IT, the ones that are happening occasionally...) but I confirm that by powering down the AE(2) playback continues. This is the first thing I will recheck when the incidence occurs again. Obviously if playback continues my conjecture (that traffic goes back to the router) will have been ruled out.

Still cannot recreate the incident. Perhaps powering on/off the AE played a key role. I will report if I notice it during the next days.

Today I had a new incidence playing Corelli's Concerti Grossi at 192kHz.

per Simon's suggestion I powered off AE2 - buffer continued. I have restarted the UpnP server (on the NAS) and keep monitoring ...

Further update: 

• powering off either/both AEs -> no effect, issue resurfaced within 5' of play
• Restarting NAS: no effect, issue resurfaced within 5' of play
• restarting NDX after power cycle of AEs: I have 45' of uninterrupted play without this proving anything if course...

 I will need to borrow another NAS for a few days but I still have the fear this is an NDX rather than a NAS issue. 

Hi again DrPo,  I thought you had satisfied yourself that the problem is not in the NDX when you played from a USB for a long period without buffering issues.  The NDX white paper block/flow diagram shows both Ethernet & USB inputting into the same streamer module & the buffers are downstream from that.   

Mike and Simon: this is a difficult issue because it happens so erratically. I have indeed tested USB and worked fine for several hours but so had my NAS for one week now... 

I know I cannot prove anything unless I get the issue with another NAS, which I don't currently have handy. In the meanwhile I will keep on reporting like a "log" of sorts in case someone else stumbles upon the same.

Hi there, I sympathise as I have a very similar issue.  Seemingly random buffering issues which in my case occurs when listening to Radio Paradise at higher bitrates.

My 'solution' after finally getting fed up with it was to use wireless connectivity on my UnitiQute in preference to a cabled connection.  I know that's strange but that has improved this.

I believe there is an underlying problem with the UnitiQute as a power cycle of that unit also also resolves things.  It's a work-around in my situation.  Network is rock solid for streaming HD from Internet so it is bizarre that 320 bit rate audio constantly fills/depletes the buffer on the UnitiQute - you can easy monitor the buffer and see what is happening.

Hi Glenw, I believe the issue you describe is a different matter and is more down to latency / TCP window sizes / Naim TCP implementation. You may find in a later firmware your TCP stability issues on specific web radio improve for a wired connection. The size of the payload in many cases becomes not as important as the timing of the flow control of the network transport protocol across the Internet.

Simon

Update today: issue persists. I have not managed to replicate it with another UpnP server yet (using Asset on my laptop as an alternative server showed no issues but this is inconclusive ...). But I have to say the only way so far to overcome the problem is to power the NDX (not the NAS) off and on. 

Hi DrPo,

As a test, would it be possible to run for some time with the NDS disconnected from AE(2) - just remove the cable from AE(2) to the switch and use the IR remote to control the system.  If the buffering issue persists then it's definitely in the NDX / switch / NAS and the AE is cleared of guilt.  If this configuration works fine, then there's something wrong with the set-up of the AE.