WHS to HD200 - Network saturation issues?

[mangriotis]

I got a (somewhat) stable version of WHS Sage Server connected to an HD200 and a client PC. I was having lots of lockups, shutdowns, resets on the HD200 due to railing my LAN, I believe (it doesn't seem to handle connection interruptions or delays gracefully at all). I can't get SA (ShowAnalyzer) to run on the WHS machine, so I am running that process on a different PC, but accessing the drive pool to do so. If I watch a show (i.e. WHS streaming to HD200), I get ~20% constant network utilization (on LAN1, as seen in task manager on WHS), and spikes to 60% or so when skipping back and forth during the show. Is this normal??

Problems arise when SA starts on the remote machine, creating additional network traffic. It looks like 50-60% constant, then railing when skipping. I had lockups often and it took some time to figure out this was causing it (I, of course, blamed HD200 firmware, WHS instability, router problems, even overheating of the HD200 ).

Additional information: I have an HDHR connected directly to a 2nd NIC (LAN2) in the WHS machine. Seems like more issues occurred when recording 2 HD streams, which resulted in ~60% utilization on LAN2 as well. Again, is this consistent with what others have seen? (Side question: The Network LED on the HDHR always blinks, is that normal when connected directly to the PC?)

Has anyone else seen this type of issue? Is there a better system design than what I have here to eliminate this kind of issue?

[Taddeusz]

What version of SageTV are you running and what HDHR driver do you have?

[mangriotis]

Latest release of Sage 6.5.9, and latest released drivers/firmware for HDHR 20081231

-- edit: I see they just released an update at Silicondust (20090215) which addresses the blinky network (link) LED, guess I will update those tonight.

[Taddeusz]

Is the main part of your network gigabit or 100Mb?

[mangriotis]

Well, the NICs in the WHS machine are both 100Mbps, as is the output of the HDHR and the HD200 (right?). The WHS goes through a Gigabit switch to the rest of the home (including the HD200, Client PC, and office PC (which is doing the SA processing)).

[Taddeusz]

Originally Posted by mangriotis

Well, the NICs in the WHS machine are both 100Mbps, as is the output of the HDHR and the HD200 (right?). The WHS goes through a Gigabit switch to the rest of the home (including the HD200, Client PC, and office PC (which is doing the SA processing)).

Yes, the HD200 is 100Mb. Since you already have a gigabit switch what I would recommend is to upgrade one of the NIC's in the WHS machine to gigabit and also upgrade the office machine to gigabit if it isn't already.

What you may be running into, though, and what I personally struggled with for a long time, is that hard drives can only have so much throughput. SA in particular uses a lot of hard drive throughput if you have it working full bore. I've since switched to comskip but even when I limited my SA to 10% CPU my HD100 playback would become stuttery. I now time limit my comskip processing so that it runs when nobody should be watching TV. I haven't had a problem with stuttery playback since.

But, as far as network throughput is concerned a single HD stream can take up to 20Mb so 20% of a 100Mb connection isn't unreasonable. Upgrade to gigabit.

[gplasky]

You definitely want a gig NIC in that server. As you can see with accessing HD files quite a bit of network traffic is created. I have a WHS with a gig NIC and right now a HD200 playing HD, a HD100 playing SD and a PC client play HD. For the server NIC the network utilization peaked at 6% and right now has leveled off between 2%-3%. The only thing I don't have going at the moment is comskip. And not only would that bring the bandwidth up a bit but it would really stress out the drives. If it is all going through a 100Mb pipe I'm not surprised it would stutter occasionally.

Gerry

[mangriotis]

I wish the problem was that it would "stutter occassionally". This problem shuts my HD200 down, severely impacting WAF ...

I have a Gb NIC I can plug into the WHS and see if it helps...

I also have made it such that SA only runs between 1 and 5 in the morning, eliminating that disk access and network traffic.

[Taddeusz]

What are the specs of your WHS machine? How old are the parts, particularly the hard drives? Are the drives you use for recording in the WHS drive pool? It should NOT be part of the drive pool. Is the drive your recordings are on formatted with 64k clusters?

[wayner]

Actually many people are starting to use a recordings drive that is in the pool, just not duplicated. S_M_E is the expert at this. I have been using drives in the pool since I went to WHS a couple of months ago with no issues but I have brand new hardware in my server (except for hard drives).

[gplasky]

As long as the drives are formatted to 64k blocks it no longer matters if they are in the pool or not. If the drives in the pool were not formatted to 64k as described in the tutorial in the forum that would be the issue.

Gerry

[mangriotis]

Originally Posted by Taddeusz

What are the specs of your WHS machine? How old are the parts, particularly the hard drives? Are the drives you use for recording in the WHS drive pool? It should NOT be part of the drive pool. Is the drive your recordings are on formatted with 64k clusters?

This is a fairly old machine: Asus A7N8X mobo, AMD 2400 (2GHz) single core processor, 1GB DDR-667, with drives pictured below (the oldest one is the system drive, which doesn't appear to have much on it if anything). All pooled drives are formatted to 64k blocks per tutorial.



I am not really using the WHS Backup drives, and will likely add one of those to the pool also. I was under the impression that the WHS horsepower did not need to be all that great, but is this stuff too old and crappy to be an effective server?

[mangriotis]

Just wanted to note that I killed the HD200 last night again by trying to fix "Verifying Port Forwarding for your router failed" messages from WHS. It seems uPnP doesn't play nice with my WRT54G (with either DD-WRT or Tomato on it). I finally configured manually but in testing that, it kicked the HD200 offline (must interrupt the connection somehow).

[gplasky]

You're using the built-in NIC? Turn off the TCP Checksum offload in the NIC settings. Also isn't there 2 NIC ports on that board? You could disable the nVidia LAN NIC and enable the 3COM one to see if that one works better. Make sure to install the driver. Best recommendation-disable them both and put a Gigabit NIC in there.

Gerry

[Taddeusz]

That may not be enough to handle things gracefully now days. What do you have your JVM heap set at? How large does your heap get? You can check by going to Settings->Help->System Information and scrolling down a bit.

[mangriotis]

Originally Posted by gplasky

You're using the built-in NIC? Turn off the TCP Checksum offload in the NIC settings. Also isn't there 2 NIC ports on that board? You could disable the nVidia LAN NIC and enable the 3COM one to see if that one works better. Make sure to install the driver. Best recommendation-disable them both and put a Gigabit NIC in there.

Gerry

I am using both NIC's - one connects to my network, and the other to the HDHR (not sure right now which is which). Both drivers are installed and up to date. I will likely add a gig NIC PCI card that I have laying around to connect to the LAN (I think this would help).

What does the TCP Checksum offload setting do for me? I can try this also. My network diagram looks something like this...

[mangriotis]

Originally Posted by Taddeusz

That may not be enough to handle things gracefully now days. What do you have your JVM heap set at? How large does your heap get? You can check by going to Settings->Help->System Information and scrolling down a bit.

My heap is default... I haven't changed anything there. I have read about this in the past, but what does changing this setting do?

When does heap size increase? During periods of lots of activity?

Right now (nothing is running at home, I fired up the Sage GUI on the WHS machine via rdp): JVM Heap Size (Used/Total/Max): 97MB/154MB/266MB

The "used" value seems to hop +-5 or so...

[Taddeusz]

Originally Posted by mangriotis

My heap is default... I haven't changed anything there. I have read about this in the past, but what does changing this setting do?

When does heap size increase? During periods of lots of activity?

Right now (nothing is running at home, I fired up the Sage GUI on the WHS machine via rdp): JVM Heap Size (Used/Total/Max): 97MB/154MB/266MB

The "used" value seems to hop +-5 or so...

You don't really need to start worrying about it unless the Total and Max values become the same. If you've ever looked at it and they are the same it means you need to increase your heap size.

[gplasky]

Originally Posted by mangriotis

What does the TCP Checksum offload setting do for me? I can try this also. My network diagram looks something like this...

This was a known bug that if you had it enabled it would slow down network access.

Gerry

[S_M_E]

WHS doesn't support multi-homing although that may not be the problem. Unsupported doesn't always mean it wont work but you might try putting the HDHR on the switch and disabling the slow NIC to see if it helps.