Problem with Network Attached Storage, HDHR3Prime, and SageDCT

[hb4]

I am trying to set up a new NAS and am able to record to it using the HDHR3 Prime, but live video freezes every second. Playback of the recording is smooth.

Memory usage of SageDCT climbs to fill available memory after about 10 minutes. Stopping, then Starting SageDCT resets memory usage and if I cancel the recording it stays constant.

With little or no networking knowledge, I am throwing my nearly octogenarian soul at the mercy of the true and valiant Sages for help. I picture those bits sent from SageDCT during the freeze periods collecting in memory and then copied in bursts to the NAS without ever being purged. Is this simply a network speed issue, or is there something more insidious going on?

Edit: Network speed test to NAS directory reported 90 Mbps

[babgvant]

See this post.

[hb4]

This afternoon I raised the Write Buffer setting in SageDCT until I could record three HD streams and watch one stream on my Buffalo Linkstation Network Attached Storage device, with the settings shown below. SageDCT memory was hovering around 70MB. I tried changing Async and RTP Buffer, but this is the only one that helped. I assume this reflects the slow speed of my LAN; I have the NAS plugged into one of two bridged ports on my Sage Server box.

I then discovered that only two of my recordings were showing in Sage so I did some more tests, ultimately raising the Write Buffer to 132K but could never repeat the steady state condition; doubling the Write Buffer for each test and cancelling the test when SageDCT memory reached 200MB of memory use.

Comments, suggestions?

SageDCT 10-23.jpg

[babgvant]

Originally Posted by hb4

Comments, suggestions?

Attachment 13593

70MB isn't bad for 3 streams.

Open resource manager and look at network utilization. If your NAS has a similar utility it would be good to take a look at that too.

Is the NAS setup as RAID 5 or 6? If so, it could be the parity calculation which is hampering write speed.

Is the network Gigabit?

How fast are the drives in the NAS?

[hb4]

For the last three tests this afternoon, it got up to 200MB and climbing. Tonight, it climbed to 525Mb after which I cancelled the three HD recordings. Memory use froze at 518Mb with no recording until I started a new recording after which memory use gradually lowered to about 50Mb.

With three streams recording and one playing network adapter speeds were:
Throughput (Send) ~55Mbps
Throughput (Receive )~40Mbps
And averaged right around 100Mb, which implies to me that I'm not running on a Gb link.

Network adapters and NAS are Gigabit but reporting 100, Bridge is reporting 1 Gb, cabling is Gigabit; The switch is 100MB; - I'd hoped that by connecting the NAS into the bridged 1Gb NIC that get the higher speed, but it doesn't look like it. The other NIC is connected to the switch.

Here's the spec on the NAS. No Raid.

2TB
Drive Interface SATA 3 Gb/s
LAN Interface
Standard Compliance IEEE802.3ab / IEEE802.3 / IEEE802.3u Standard
Data Transfer Rates 10 / 100 / 1000 Mb/s
Connector Type RJ-45
Number of Ports 1
USB Interface
Standard Compliance USB 2.0
Connector Type A type
Number of Ports 1
Data Transfer Rates Max: 480 Mb/s (High Speed Mode)
Protocol Support
Networking TCP/IP
File Sharing CIFS/SMB, AFP, HTTP/HTTPS, FTP
Management HTTP/HTTPS
Time Synchronization NTP
Other
Dimensions (W x H x D) 1.78 x 6.89 x 6.15 in
Weight 2.5 lbs
Power Consumption (Watts) 17 W (Average)
Power Supply AC 100-240V 50/60 Hz

[babgvant]

Originally Posted by hb4

Network adapters and NAS are Gigabit but reporting 100, Bridge is reporting 1 Gb, cabling is Gigabit; The switch is 100MB; - I'd hoped that by connecting the NAS into the bridged 1Gb NIC that get the higher speed, but it doesn't look like it. The other NIC is connected to the switch.

Is the NAS directly connected to the PC via Ethernet then?

[hb4]

Is the NAS directly connected to the PC via Ethernet then?

Yes. I have two ethernet ports in the PC - a Realtek on the MB and a Dynek card, both with Gb capacity (1/100/1000), bridged via W7 software. A diagnostic program reports 1Gb link speed for the bridge but 100Mb for each port. The NAS is connected to the MB port, and the LAN is connected to the Dynek. The LAN consists of a wireless access point, a router, and a bunch of PCs. The NAS is accessible via any of the PCs on the LAN.

EDIT> I'm running the test this morning and now the Realtek reports 1GB link. I set it up to record three HD streams and watch one; choosing HD streams that wrote the largest files to the NAS. Comskip was working on one of the streams. SageDCT memory grew to 200Mb at which point I stopped and started the service via SageDCT config. The Realtek card hovered around 100Mb for this part, and it seemed that when the card throughput was high, SageDCT memory grew and vice versa. Once the service restarted, Comskip was no longer working any of the streams and SageDCT has been holding at around 45Mb for the last few minutes. The Realtek card is now reporting around 95Mbps. Seems like Comskip caused SageDCT memory to grow; at least those are the parts of the elephant I can see.

No changes made to Network topology or NIC card setup, W7, SageDCT, or Sage since last night.

[mikejaner]

bridged via W7 software. A diagnostic program reports 1Gb link speed for the bridge but 100Mb for each por

This sounds fishy. I personally would remove any complexity to your config, like the bridging and test from there.
My experience with NAS boxes has been an interesting one. I have done tests with Freenas, Ubuntu, Openindana, Server 2008, and Server 2012. The interesting part is I noticed a pattern with Windows Vs. Others. On both 2008, and 2012 with a Software RAID setup, I have no problems recording over the network, from my Win7 SageTV box. If I try to record to a Linux Software RAID, or OpenIndiana exposed via CIFS/SMB, I get studdering and failed recordings. Playback is fine for existing recordings etc. All of these shares on the *nix systems don't have any issues with other things, like file copies etc..., but when it comes to real time video recording, there is problems.

[babgvant]

Originally Posted by mikejaner

The interesting part is I noticed a pattern with Windows Vs. Others. On both 2008, and 2012 with a Software RAID setup, I have no problems recording over the network, from my Win7 SageTV box. If I try to record to a Linux Software RAID, or OpenIndiana exposed via CIFS/SMB, I get studdering and failed recordings.

Parity calc is expensive. IMO, for most use cases RAID 5/6 is poor choice and provides a false sense of data security. If you want proper real time data protection use RAID 1 (or 10), otherwise JBOD.

[mikejaner]

Originally Posted by babgvant

Parity calc is expensive. IMO, for most use cases RAID 5/6 is poor choice and provides a false sense of data security. If you want proper real time data protection use RAID 1 (or 10), otherwise JBOD.

My opinion is different. In the case of today's processors, it's not expensive at all. I can write a steady 100+MBytes/Sec to a RAID 5 array on both types of systems, with exception to Server 2012 (but that's another story for a Storage Pools Discussion). The processor cost is ~5% when recording TV on multiple tuners with SageTV (40-60MBits a second write, plus one or two Extenders Watching/Reading from the same drive).
Either way, I wanted to test like setups, so I did RAID5 across the board since I started with a RAID5 way back in the mid 2000's when drives were more expensive.

[babgvant]

Originally Posted by mikejaner

My opinion is different. In the case of today's processors, it's not expensive at all.

Semantics somewhat. Something doesn't become inexpensive just because you're rich.

Parity calc feels cheap because you are doing it on a powerful PC with more than enough resource headroom. When it's done on an underpowered NAS box (many are ARM SOC optimized for power consumption), the story is much different.

Originally Posted by mikejaner

Either way, I wanted to test like setups, so I did RAID5 across the board since I started with a RAID5 way back in the mid 2000's when drives were more expensive.

The problem with RAID 5/6 is that when you have a drive failure it's quite likely that you will have another failure before recovery is complete. This risk increases as the size of the drives in the pool goes up.

[mikejaner]

Semantics somewhat. Something doesn't become inexpensive just because you're rich

AMD Athlon II X2 $60 two years ago, so, not rich at all.

Parity calc feels cheap because you are doing it on a powerful PC with more than enough resource headroom. When it's done on an underpowered NAS box (many are ARM SOC optimized for power consumption), the story is much different.

Correct, those Dedicated NAS boxes are all over the place in quality/performance.

I didn't read his second post that he was using a Buffalo NAS, I thought from his first post it was a PC based NAS.

My argument is, it's not expensive as a software RAID on any Mid to High end PC you would have purchased in the past 5-7 years (What I should have said instead of "todays processors"). In my case it was a spare motherboard and case populated with $25 of ram and a $60 processor two years ago.

The problem with RAID 5/6 is that when you have a drive failure it's quite likely that you will have another failure before recovery is complete. This risk increases as the size of the drives in the pool goes up.

Interesting article. I am glad that has never happened to me though. I have only had a up to 6TB array over the years though. I have had a few drive failures over the years, but never a failure during the rebuild.

[babgvant]

Originally Posted by mikejaner

My argument is, it's not not expensive as a software RAID on any Mid to High end PC you would have purchased in the past 5-7 years (What I should have said instead of "todays processors"). In my case it was a spare motherboard and case populated with $25 of ram and a $60 processor two years ago.

OK, but again that just changes the feel because in relative terms the PC is rich (i.e. has overly abundant resources) - not the fundamentals.

I agree that PCs make good NAS for this reason, and many current gen stand-alone devices are essentially PCs w/ x86 CPUs.

[mikejaner]

Originally Posted by babgvant

OK, but again that just changes the feel because in relative terms the PC is rich (i.e. has overly abundant resources) - not the fundamentals.

I agree that PCs make good NAS for this reason, and many current gen stand-alone devices are essentially PCs w/ x86 CPUs.

Gotcha BTW love SageDCT!!

The spirit of my original post is that I suspect a Windows to *nix (buffalo) SMB issue or a network issue. There is just something weird to me about how poorly streaming video gets written in a non MS share.

[hb4]

Just an update; I installed a Netgear 8-port gigabit switch and connected the NAS, tuners, ReplayTV (Poopli/WIRNS for all those shows that my Sage setup misses), wireless access point, Sage Server, and cable from router.

Didn't have time to do much testing before I left town for a week, but did notice a few things after I started three HD streams and viewed one:

The NIC port reported over 100Mb for the first time.
With three comskip operations allowed, memory started to climb towards 400MB, but after I allowed only one it never got above about 50Mb. I'll do some more testing when I get back.

I think my setup probably represents the level of technology to which an average non-IT professional user might take SageTV, i.e, consumer level NAS, switch, and router and W7 - rather than a Server and PC-based NAS and router. I'm using the HDHR3, an HDHR, HDPVR (that is fussy), and OTA. Overall, I am amazed at how well Sage does with all of this and really appreciate the folks on this Board that contribute their expertise. The only thing I can see that would make me switch is if Comcast went to copy once on all channels and forced me to use the HDPVR for everything.

The Prime and SageDCT is the best!