First time poster so apologies if posting to the wrong thread.
I have latest public LE (8.0.2) running on my Odroid-C2.
Yesterday I notices that the transfer speeds from KODI (playing back a video from the NAS over CIFS) range at about 10-12Mbps which is fairly sufficient for the most content i have with exception of some videos with higher bit rate. As I am running my NAS in an "off-site" location and connect to it over VPN I did not pay much attention at first.
Yesterday I performed some proper measures and the results are more than curious.
The testing setup is as follows:
Odroid-C2 with this image installed
LE 8.02 running from an 16GB eMMC
C2 connected directly to a router/firewall via ethernet (1000Mbps full duplex)
adnacedsettings.xml set to most aggressive settings possible
<advancedsettings>
<cache>
<buffermode>1</buffermode>
<memorysize>0</memorysize>
<readfactor>20</readfactor>
</cache>
<pvr>
<minvideocachelevel>65</minvideocachelevel>
<minaudiocachelevel>65</minaudiocachelevel>
<cacheindvdplayer>true</cacheindvdplayer>
</pvr>
<videolibrary>
<importwatchedstate>true</importwatchedstate>
</videolibrary>
</advancedsettings>The network utilization is measured on the router port.
Scenario 1: Play back a movie from the video library stored on the NAS (CIFS)
Speed peaking at 12.1Mbps
Scenario 2: copy a file from the NAS to /storage/.kodi/temp/
Speed peeking at 35Mbps (the maximum speed ov the VPN tunnel; see above)
Scenario 3: iperf3 test against the NAS (Win2016)
Speed peeking at 25/35Mbps (exactly as per the specifications of my ISP resulting at 100% utilization of the line)
Scenario 4: iperf3 test against a Linux machine on the same (wired, gigabit) network
Speed peeking at 780Mbps, which is a result I am comfortable as the opposing Linux test machine has some performance issues.
Scenario 5: copy a file from the NAS to /dev/null/
Speed similar to Scenario2
After those tests I am confident I can conclude KODI as an app is the bottleneck and my question is is there I as an user can do to make it use as much of the bandwidth as possible?
Any ideas of further testing and tinkering are more than welcome
