HD300 User interface lag

[Wrx4me]

I recently switched from Directv to Fios and surprised myself by how easy the transition was to complete. I have one Motorola qip7100-2 receiver connected to an HDPVR for HBO/Cinemax using FireWire channel changing. I also have 2 HDHome run primes running cablecards.

Everything works beautifully, except that on the HD300's the UI is sticky if you know what I mean. As you scroll through the guide, there is a noticeable lag while a show is being watched but everything returns to normal if playback is stopped. I have no issues with my one PC client.
Anyone have any ideas on what I should check for on the extenders? I would appreciate any suggestions. Thanks!

[Wrx4me]

Well I guess I must have a unique problem. No suggestions after 4 days.
I do know the issue is caused by the UI animation when video is being displayed. Turning off the animation fixes things but of course scrolling is no longer smooth and fluid.

[gveres]

Are you using the Gemstone UI? If not, I would change over to it. I have noticed that with the change from Diamond to Gemstone, the UI performance improved a lot. I am getting pretty reasonable performance on my HD100 box using Gemstone where it was completely unbearable in the past.

[tmiranda]

The usual cause of this is running out of java heap space. In the default STV (from one of the laggy extenders) go to Setup->System Information and scroll down until you see "JVM Heap Size". Post your values.

[Wrx4me]

@ Gveres I will look into Gemstone if I cant get an acceptable resolution to this issue, thanks.

Tmiranda my java heap numbers are (used/Total/Max) 492MB/825MB/1038MB.
I am also running Java version 1.6.0_29 on the server, I don't know if that has any bearing on things. Thanks.

[tmiranda]

If the three numbers are all close to the same it usually means you are maxing out your heap. Yours look fine.

One oddity is that you have allocated 1038MB which is a bit over 1GB. On some systems once the heap goes over 1GB Java crashes or Sage does not work right. It wouldn't hurt to change your heap to 1024MB just to make sure this is not the issue.

[Wrx4me]

My Java heap is set to 1024 but reports as 1038. Also when I set it to 300 it shows in Sage as 779MB. Maybe because I have 4GB memory on a 32bit Win7?

[wbarber69]

Originally Posted by Wrx4me

My Java heap is set to 1024 but reports as 1038. Also when I set it to 300 it shows in Sage as 779MB. Maybe because I have 4GB memory on a 32bit Win7?

Mine reports as 1038, when reg is set for 1024 here too.

[PhilH]

Originally Posted by wbarber69

Mine reports as 1038, when reg is set for 1024 here too.

Same here also...

[Wrx4me]

I guess that's how Sage interprets 1024MB.

Next step for me will be to pull my unused Colossus which is still in my server, don't know if that has any bearing on my issue but the Colossus always gave me problems when I used it.

I was also thinking of swapping out the switch that my server and one of the HD300's is connected to but the other HD300 is connected to a different switch upstairs so I don't know how that helps the situation.

The one constant is that everything is on gigabit ethernet with not lag in video play except when browsing the UI with video in the background.
Only change has been ripping out my HDPVR's and replacing them with HDHomerun Primes and going from Directv to FIOS.

I guess the default UI doesnt like the FIOS signal.

[Wrx4me]

I finally found the cause of my problems after searching the forums and finding the following thread. http://forums.sagetv.com/forums/show...eceive+buffers

My Sage server had always been connected to a 100Mbps switch until I ran Gigabit Ethernet upon getting FIOS.
I then put the server on a gigabit switch and it connected to the network at 1Gbs. This change from 100Mbs to 1Gbs is what was causing the laggy UI on the extenders, clients work fine.
As soon as I forced the server NIC down to 100Mbps, my UI became smooth as silk. I turned the NIC backup to Auto Negotiate or 1000Mbs full and all the lag returned.
I guess I have to run my server at 100Mbs until I can figure out how to run it at gigabit speeds without the extenders having an issue. But for now all is well again.

[drewg]

Originally Posted by Wrx4me

This change from 100Mbs to 1Gbs is what was causing the laggy UI on the extenders, clients work fine.

Try playing with the ethernet flow control (also known pause frames) on the server when it is connected at 1Gb/s. The problem is that the extenders max out at 100Mb/s, but the server is running at 1Gb/s. So you are sending from a fire hose to a garden hose, if you will. When the server sends more packets in a burst than the switch can buffer, then it will drop them. When things get dropped, they need to be re-sent, and you get lag. Depending on the switch (and the NIC), enabling / disabling ethernet flow control can change when/where things get dropped, and can possibly solve your issue. I had a similar issue a few years ago when I replaced an HD100 with an HD300.

Disabling pause on the server solved it for me. This was on an AMD server with some $0.99 on-board NIC. When I upgraded the server to one with an Intel NIC, the problem went away entirely. My theory was that the HD100 did not support pause, but the HD300 did. And once it started to send pause frames, it triggered a bug in the pause frame handling on the server. In your case, I could imagine that your old switch did not support pause, but the new one does.



Drew

[Wrx4me]

Good point Drewg. I am using an on board Realtek NIC
I will try your suggestion later when the days recordings are done. Thanks.

[MattHelm]

Originally Posted by drewg

.... This was on an AMD server with some $0.99 on-board NIC. When I upgraded the server to one with an Intel NIC, the problem went away entirely. My theory was that the HD100 did not support pause, but the HD300 did. And once it started to send pause frames, it triggered a bug in the pause frame handling on the server. In your case, I could imagine that your old switch did not support pause, but the new one does. ...

Cheap switches can also cause the same issue. They can ignore the pause, or pause ALL traffic thru them. I had issues with PC to PC file transfers because of this, until I changed to a mid range (~$80) HP switch.

[Wrx4me]

Enabling Flow Control caused a never ending spinning icon so I disabled it and backed the NIC down to 100Mbs.

[drewg]

Originally Posted by MattHelm

Cheap switches can also cause the same issue. They can ignore the pause, or pause ALL traffic thru them. I had issues with PC to PC file transfers because of this, until I changed to a mid range (~$80) HP switch.

And at least some of the HP 10GbE switches take a pause frame as a signal to drop everything in their transmit buffer. Which can be a bit of a bummer..

Drew

[MattHelm]

Originally Posted by drewg

And at least some of the HP 10GbE switches take a pause frame as a signal to drop everything in their transmit buffer. Which can be a bit of a bummer..

Drew

I hope they plan on fixing that?????

[drewg]

Originally Posted by MattHelm

I hope they plan on fixing that?????

Not AFAIK. To them, it is a feature, not a bug. They protect themselves from blocking in their core crossbars, and don't technically violate the spec. What they're really trying to do is to protect themselves from a NIC in a crashed host that keeps emitting pause frames forever because the host has crashed & cannot supply more rx buffers. Eventually the switches xmit buffers would fill, then they'd back up & block the crossbars, and the whole switch could be fubar. We actually have a watchdog in our NIC's firmware to turn off flow control if the host has not responded in a "reasonable" amount of time. I only wish other vendors did the same, so that switch vendors did not have to be so paranoid. We seem to have wondered far from the original topic though.. Back to the original topic:

I second the suggestion to try a new switch. Or, if possible, use a higher quality NIC in the server. Intel is generally a good choice (note: I do not work for them. )

Drew

[Wrx4me]

I plan to pick up a new NIC for the server but in the meantime I am now running the server at a gigabit. The lag is not as bad anymore and I can live with it for now.

The 100mb speed was too slow considering I have 9 tuners, 8 of which are networked.

Yesterday I had 6 recordings going at once on the 2 HD Homerun primes and my network utilization on the server was 146Mb so gigabit gives me more headroom.

[Wrx4me]

I replaced my on board NIC with an Intel gigabit PCIE NIC and all the laggy performance has disappeared.
Everything is as smooth as butter once again
It looks like my onboard NIC is on the fritz.