long range wifi pinapple. (throughput speeds 7.2mbps)

[i8igmac]

So, i have been playing with some long range repeater like devices... 2000 feet and several walls to go threw, rebroadcasting this network access in your own home. With permission...

What kind of download speeds are you capible of repeating...

Durring the best conditions, the first day I setup my network I seen 7.2mbps from testmt.net/dl-5000

Raspberry pi, awus036nha, also a alfa r36 router

[cooper]

That testmt site does nothing (useful) for me. What throughput are you getting when you attach a laptop to the router on one side and you let the pi suck in a file from the laptop, diverting the data stream straight to /dev/null ?

[digip]

Wifi signals will always depend on your circumstances, from line of sight, interference, antennas, router wifi capabilities, to the modem's wired speeds which you pass through before hitting the Internet, which will all come into play.

The home I live in now I am bridged from one end of the house to the other end of the house via 802.11N only, while downstairs my router is setup for everyone else's wifi to connect through me. My internet speeds from places like http://www.speedtest.net/ and http://www.speakeasy.net/speedtest/ are generally the same at roughly 28 down and 10-11 up(via my router over wifi only). I'm about 40 feet away and have several walls and metal objects in the home to go through, but I get about 50 down and 15-20 up on wired from the upstairs router itself. Throughput from router to router ranges from >54 - <150+mbps over wifi(The Asus tells me my connectivity speed to the other router) but will dumb down my actual internet speeds when things like an 802.11b device connects to drop speeds a bit even with the mimo antenna, since the N one is used to bridge us to the upstairs router and also other N devices, my router will get slower when other people aren't using N only.

This isn't bad considering the upstairs router is slower than my own downstairs one to begin with. Upstairs is a Cisco E2000 and downstairs I have an Asustek RT-AC66U. When I was on just the Asus and my own modem at my old place, I used to get 100+ down and about 45-50 up, wired broadband and even on wireless devices that used N only(and I didn;t use the B/G side of the router), to the internet(not home connectivity which was way faster locally) but I'm not living in the same place nor using only my equipment and modem to connect to the internet right now.

By the way, the ops link should be testmy.net not testmt(.)net - Try http://testmy.net/SmarTest/down and it will automatically adjust speeds, but I don't trust their results since you don't know where the server location is in refernce to you.

Edited March 15, 2015 by digip

[cooper]

One thing you could look at is the load on the Pi as it's sucking its data in. The processor on there isn't exactly a sprinter so you may have to tweak things a bit to get better throughput. Check out my spoiler in this post where I optimize things for Linux on my PcDuino3 Nano - they may very well benefit the Pi too.

[i8igmac]

One thing you could look at is the load on the Pi as it's sucking its data in. The processor on there isn't exactly a sprinter so you may have to tweak things a bit to get better throughput. Check out my spoiler in this post where I optimize things for Linux on my PcDuino3 Nano - they may very well benefit the Pi too.

All this is starting to make sense, turns out the pi eth0 and usb share the same buss or something, they all fight over throughput... cpu load effects throughput speeds as well as i noticed some of my monitoring tools cpu usage was high and had a effect on network speeds, causing false possessives

i discovered that this alfa r36 router has lost its passthrough speeds, something about alfa hardware is starting to tick me off...

so, i replaced the alfa r36 router with a fresh ddwrt router that suppports wireless n descent ram etc... i got my speeds back ! conditions are not optimal as there currently ten devices, torrent downloads and i can get 3.5 mbps at testmy.net/dl-5000 durring these current conditions

Edit. Looking for some performance tweeks, currently a default install of kali dist-upgraded... any tips, found some good tutorials on google already...

Edited March 16, 2015 by i8igmac

[i8igmac]

There is a tweak im interested in, it seems that the ram is split into separate parts, cpu/gpu memory... there are gpu_mem tweaks that could lower the memory saved for the gpu

gpu_mem=16

http://elinux.org/RPiconfig

"gpu_mem GPU memory in megabyte. Sets the memory split between the ARM and GPU. ARM gets the remaining memory. Min 16. Default 64"

at the moment, no torrent downloads running. some what idle network activity

free -m

             total       used       free     shared    buffers     cached
Mem:           437        192        244          0         32        124
-/+ buffers/cache:         35        401
Swap:            0          0          0

I am running kali, so maybe this tweak is not necessary? i dont know how to confirm this tweak... i guess i need to download some files from git /boot directory then chmod 555 ?

[cooper]

Rather than looking at the output of 'free', check out /proc/meminfo

To give you some idea of what's possible, this is what I'm providing to the kernel on my nanos, yielding roughly an extra 25MB of memory:

console=tty0 hdmi.audio=0 disp.screen0_output_mode=1280x720p50 console=ttyS0,115200 root=/dev/mmcblk0p2 sunxi_g2d_mem_reserve=0 sunxi_fb_mem_reserve=16 consoleblank=0 rootwait panic=10 zswap.enabled=1

In this, sunxi is the device type (AllWinner & friends). The thing to remember is that not only memory itself is limited - so is memory bandwidth. My devices run headless unless I can't get on them via the network, so by setting the screen output mode to the lowest HDMI mode accepted with a low refresh rate this chunk of shared memory can be smaller and accessed less frequently. Setting g2d memory reserve to 0 basically means I can't have GPU acceleration of stuff like video which is a-ok for me with this usage scenario. Setting fb mem reserve is similar to the setting you're applying I guess. Setting consoleblank to 0 (=off) means there doesn't have to be a timer thread that needs to keep track of the screen needs to go to black. Finally there's zswap. Here's an LWN article on zswap. Even without a swapfile it can sometimes sort-of increase the effective size of the available memory at the expense of a bit of CPU. This one will also cost you memory bandwidth, but since I feel memory is a more precious commodity on my boards than CPU cycles and, in case of creating a temporary swapfile on the SD card, the same can be said for the writes on my SD card, I think this is justified.

[i8igmac]

Are these all options I can write to a file in the boot directory? Do I need any other files for these options to work for a kali installation?

Raspbian I believe some files exist in the boot folder, I have read about wget these files from there git repo

[cooper]

Perhaps you can specify the contents of /boot as I don't have a Pi, don't run Kali and actually never even seen a system with that OS on it...

[i8igmac]

The boot directory on my kali pi is empty, I have read about downloading the boot directory from raspbian git repo. Then I should be able to make use of the options.

[cooper]

Try to mount /boot and I think you're going to find some files there.

[i8igmac]

http://i.imgur.com/imJaqVc.jpg

http://i.imgur.com/ihhLcfE.jpg

A little off topic, look what i pulled off a enduction cooktop, huge heat sink and brushless 12v fan! ill bet i can fit 3 pi's on this hunk of aluminum... overclocking is now a option...

Its about 12 inch's long, over 1 inch thick with a tupe for the air to pass through...

Ill imbed the images when i get home...

Edited March 18, 2015 by i8igmac

[cooper]

Overclocking the original Pi.

Overclocking the Pi 2

[i8igmac]

I hope this better explains my current network configuration... my attempts to get the best throughput repeater speeds

/etc/rc.local

echo '1' > /proc/sys/net/ipv6/conf/all/disable_ipv6
iptables --table nat --flush
iptables --table nat --delete-chain
iptables --table nat --append POSTROUTING --out-interface wlan1 -j MASQUERADE
iptables --append FORWARD --in-interface eth0 -j ACCEPT
echo 1 > /proc/sys/net/ipv4/ip_forward
echo 'interface=eth0' > /etc/dnsmasq.conf
echo 'dhcp-range=192.168.96.50,192.168.96.150,12h' >> /etc/dnsmasq.conf
ifconfig eth0 192.168.96.1 up
dnsmasq
wpa_supplicant -Dnl80211 -iwlan1 -c/root/golf7376.conf -B

exit 0

Edited March 18, 2015 by i8igmac

[digip]

Have you compared the speeds of a wired desktop on the same network using the same speed test site as well as other wifi devices at the home(yours and where the exit hop is beofre the internet), along with them connecting via the repeaters at various places in the line of site? I think in the case of using an external site for speed testing of an internal long range setup, you're at most only testing the exit from your Pi to the Internet via this setup but not fully looking at the pipe between you and the internet first before expecting better speeds for which, you may already be getting normal speeds with the current network mesh in place.

It may be all you'll get and relative to speed of the network it's on more than what device you're using to get online with. I can understand other devices probably getting better speeds and throughput, but average out multiple devices in the same scenario, as well as take your Pi to the base site where the wifi is without repeaters and see how it fares both locally wireless and wired at the source of your internet and see what happens.

[xmx1024]

It seems you have made a typo in your URL, OP.

http://testmy.net must be what you meant.

[i8igmac]

edit: installed kali 1.1.0 lastnight...

ls /dev/mmc*

/dev/mmcblk0 /dev/mmcblk0p1 /dev/mmcblk0p2

mount /dev/mmcblk0p1 /boot/
ls /boot

bcm2708-rpi-b.dtb bootcode.bin fixup_cd.dat kernel7.img overlays start_x.elf
bcm2708-rpi-b-plus.dtb cmdline.txt config.txt fixup.dat kernel.img start_cd.elf
bcm2709-rpi-2-b.dtb COPYING.linux fixup_x.dat LICENCE.broadcom start.elf

all your over clocking options for config.txt are in the links i provide...
i just wanted to share... i have not tested but here it is... insert your kali installed sdcard into your labtop, you will see 2 partitions... /dev/sdc1 and /dev/sdc2, you will notice some files exist in sdc1... bootcode.bin and start.elf, write your config.txt here... the links i provide will show all options for config.txt


kali /boot folder is empty
https://github.com/raspberrypi/firmware/tree/master/boot
http://www.raspberrypi.org/documentation/configuration/config-txt.md usefull document for boot options
http://elinux.org/RPiconfig more info

As it's an embedded platform, the Raspberry Pi doesn't have a BIOS like you'd find on a conventional PC. The various system configuration parameters, which would traditionally be edited and stored using a BIOS, are stored in an optional text file named config.txt. This is read by the GPU before the ARM CPU (and Linux) is initialised; therefore it must be located on the first (boot) partition of your SD card, alongside bootcode.bin and start.elf. This file is normally accessible as /boot/config.txt from Linux and must be edited as root; but from Windows or OS X it is seen as a file in the only accessible part of the card. If you need to apply some of the config settings below, but you don't have a config.txt on your boot partition yet, then simply create it as a new text file.

Edited March 25, 2015 by i8igmac