TechNazgul RSS

Tuesday, February 8, 2011

NFS vs. SMB throughput for HD media streaming using the NETGEAR XAVB5001 500 Mbps Powerline Adapter



NETGEAR XAVB5001 500 Mbps Powerline Adapter Review:


XAVB5001After recently dropping my satellite subscription and moving to MythTV, I needed to find a way to get HD video streams from my antenna and HD Homerun tuner to a TV in the house without wired internet.  Wireless (G or N) didn’t prove to be reliable or fast enough, so the Netgear 500 mbps (theoretical) powerline adapters seemed like a good bet.

Testing was performed in a ~3-year old house and the 2 adapters were several rooms apart from one another on the same floor.  With these adapters, performance can typically vary significantly based on wiring within the house, so YMMV.

As a preview of the results, while they didn’t achieve anywhere near the advertised 250 mbps (~31 MB/s) in each direction, they did perform much better than older generation powerline adapters I’ve used.  When I plugged in the adapters, the Netgear software reported that I was getting ~220 Mbps in each direction, which was encouraging, but ultimately not even close to accurate.

In addition to testing the adapters, I also wanted to test the efficiencies of different network protocols to see what would work best for streaming HD content to my XBMC / MythTV box.  Below are the results.

Testing was done using commands like the following on a Ubuntu Linux machine that tests reads and write speeds to network-mounted disk. 

Write: time dd if=/dev/zero of=/mnt/storagesmb/testfilesmb1 bs=16k count=16384
Read:  time dd if=/mnt/storagesmb/testfilesmb1 of=/dev/null bs=16k

Since the focus of my testing was primarily to verify read speeds from a media frontend to a backend storage server, I did not do significant write testing.

Disk Read/Write Test Results:

Protocol Test Type File Size Rsize (nfs) Throughput
SMB Write 256MB na 03.8 MB/s
SMB Read 256MB na 03.9 MB/s
NFS v3 Write 256MB na 05.4 MB/s
NFS v3 Read 256MB 16,384 09.3 MB/s
NFS v3 Read 256MB 32,768 10.8 MB/s

Next, a separate set of tests was done using iperf between two Ubuntu Linux machines.  The server resides on a large RAID array capable of reads and writes upwards of 500 MB/s, so disk speeds were not a limitation in this testing. 

In theory, the iperf test should remove most inefficiencies related to network protocol and show raw throughput. As it turns out, the results are very similar to the NFS testing above, which speaks well of the NFS protocol.

iperf results:

***ONE DIRECTIONAL TEST
------------------------------------------------------------
Client connecting to 192.168.1.111, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[  3] local 192.168.1.182 port 42245 connected with 192.168.1.111 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  90.8 MBytes  75.9 Mbits/sec (9.49 MB/s)


***BIDIRECTIONAL (SIMULTAENOUS) TEST
------------------------------------------------------------
Client connecting to 192.168.1.111, TCP port 5001
TCP window size: 55.0 KByte (default)
------------------------------------------------------------
[  5] local 192.168.1.182 port 60753 connected with 192.168.1.111 port 5001
[  4] local 192.168.1.182 port 5001 connected with 192.168.1.111 port 59780
[ ID] Interval       Transfer     Bandwidth
[  5]  0.0-10.0 sec  57.4 MBytes  48.0 Mbits/sec (6.00 MB/s)
[  4]  0.0-10.1 sec  43.0 MBytes  35.7 Mbits/sec (4.46 MB/s)

Combined => ~83.7 Mbits/sec (10.46 MB/s)



SMB vs. NFS Conclusion:
While this is probably no surprise to many, NFS is vastly more efficient.  Using the same commands and testing read speeds with a 32k rsize, nfs outperforms smb by ~280%

The difference was noticeable on XBMC performance as well.  When my network shares were mounted using SMB, I was unable to smoothly stream high bitrate uncompressed 1080p bluray rips (i.e. Avatar).  However, when using NFS, it played absolutely everything I could throw at it.

Netgear XAVB5001 Conclusion:

Bottom line, if your wiring is fairly modern and free of interference, while these adapters fall far short of their advertised 500 Mbps speeds, they should be more than capable of streaming your uncompressed HD content if you can serve your files via NFS.  Even if you are stuck using SMB (Windows network file sharing), most of your content will work without issue, but high bitrate uncompressed bluray will likely not play 100% smoothly.


2 comments:

  1. What kind of latency do you get with these? I've been wondering lately if my wireless connection (using a Linksys E4200, connecting on the 2.4 GHz band using wireless N (that band is also carrying a wireless g connection) would use the 5GHz band, but the adapter I bought is broken) is causing higher latency for gaming.

    ReplyDelete
  2. I don't remember exactly, but do recall that it was higher than when using my wireless network. Maybe 30 ms, but really I'm just guessing. I don't have them plugged in any longer so unfortunately I don't have an easy way to give you exact numbers.

    ReplyDelete

Followers

Facebook