Bug #463

Peformance tune wifi network

Added by Hammel almost 2 years ago. Updated 9 months ago.

Status:ClosedStart date:02 Aug 2015
Priority:UrgentDue date:
Assignee:Hammel% Done:

100%

Category:03 - Linux Kernel
Target version:0.11.0
Severity:05 - Very Low

Description

Is there anything I can do to improve performance of the wifi network? Any kernel parameters that can be tweaked?

I get dropped playback of videos if I have more than one player at a time (player and tablet, for example).

Associated revisions

Revision 8e0a5286
Added by Hammel 9 months ago

RM #463: Updates for new wifi adapters and general cleanup.

Revision 33e4f659
Added by Hammel 9 months ago

RM #463: Add network performance measurement tools.

Revision 609f40b0
Added by Hammel 9 months ago

RM #463: Add wifi tuning parameters to bootup, some of which are disabled since they may not really be needed.

History

#1 Updated by Hammel over 1 year ago

  • Priority changed from High to Urgent
  • Severity changed from 01 - Critical to 05 - Very Low

#2 Updated by Hammel 9 months ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 50

I've tuned a Model 2 server and a Model B+ as much as I can. I only improved performance slightly. The Model 2 does slightly better. The Model B+ is in a box with a pico projector. Heat in that box may play a role in reducing throughput.

I've verified that the wifi plays a big role in the poor performance of the videofe on the player (Model B+) by connecting the eth0 ports to a hub. Through the wired connection I get 12MBytes/s throughput but with wifi I get only 48KB/s.

I've got some alternative wifi adapters on order, based on reviews for RPi dongles. I'll try those. For the demo at Maker Faire I'll also configure the eth0 ports and take the hub.

#3 Updated by Hammel 9 months ago

I ordered the following and they arrived today:
  1. 2 - Panda Ultra 150Mbs Wireless N, Model PAU03
  2. 2 - Panda Wireless 300Mbs N Model PAU05
  3. 2 - Wi-Pi
  4. 2 - Edimax N150 Nano USB Adapter

All of these were recommended by various sites.

I'm going to get specs on them - like which driver they use - by testing them on my desktop first.

#4 Updated by Hammel 9 months ago

  • % Done changed from 50 to 70

I tried the Panda 300Mbs in the player first. I immediately got a jump to 3.x+Mb/s. Huge difference. But the player was still very slow starting and reacting to browsing the video list.

So I checked the file copy of the posters manually. They still take about 24s each. That's because, at least in the test directory of posters, they are over 2MB each.

So the next step is to try and resize and png-crush them. I'm actually using optipng instead of pngcrush because initial tests show a slightly higher reduction with optipng. And I'm scaling them all to 400x400. The reduction is dramatic. The test poster went from 2313972 to 168509 bytes and the quality is still fairly high. Now I've written a script to do the same thing to all the files in the posters test directory and will retry starting the player using the server's poster test directory. I did a test to see how long it takes to transfer the new poster and it was nearly instantaneous.

Assuming this improves the situation significantly it will mean adding an option to VideoLib to resize, convert (if not in png format) and crush the images. That process actually takes a while for each image. So I'll have to make it optional and pop up a warning note that it will take a while.

#5 Updated by Hammel 9 months ago

  • % Done changed from 70 to 90

Bingo. With the new wifi adapter in the player I'm able to browse the reduced sized video posters much faster in videofe. For grins I started a video on the player, a video on the server (a Model 2) and a video on my Samsung Galaxy Tab. They are all playing at the same time (I have video to prove it) with no lag, jumps or stutters.

So I can close this issue after I open one for updating videolib with the poster compression support.

#6 Updated by Hammel 9 months ago

  • Status changed from In Progress to Closed
  • % Done changed from 90 to 100

Compression tracking issue opened.

Closing this issue.

Also available in: Atom PDF