Frenzy Virtual Traffic Shapping September 25, 2012

Walkthrough

Hopefully this will be a relatively simple setup on configuring a transparent bridge to perform traffic shaping & manipulation. This will allow you to inject artificial latency, bandwidth limits, and packet loss across multiple VirtualMachines for testing.

Obtain Frenzy.

This uses the bootable FreeBSD distrobution called Frenzy. The latest, as of writing, is 1.4 which is based off of FreeBSD 8.3 Stable.

    http://frenzy.org.ua/en/index.shtml

    wget ftp://ftp.frenzy.org.ua/pub/Frenzy/1.4/iso/frenzy-1.4-lite-en.iso

You will of course now need to get that into your shared storage within vCenter.

Setup virtual networking

You will want to create two virtual standard switches on your ESXi server. One vSwitch will have no uplinks and the other will have one.

    We will call these two vSwitches vss_Isolated and vss_Lab

Add/Change all the network adapters for all the virtual machines that you want to manipulate to the vSwitch that has NO uplink.

    Should have been called vss_Isolated
        
    note: You can see a screenshot of how I have set it up within this repo.

Create Frenzy Virtual Machine

Create a small virtual machine with roughly 256M of memory and no hard disk. Create two network adapters. Attach the first adapter to vssIsolated and next adapater to vssLab. Attach the Frenzy iso to the cdrom and ensure it connects at boot up. Power on the newly created Frenzy Virtual Machine.

Configure Frenzy for traffic manipulation

Once Frenzy finshes booting up it will drop you right to a root prompt.

    note: em0 should be vss_Isolated and em1 should be vss_Lab

create bridge

    ifconfig bridge create
    ifconfig bridge0 addm em0 addm em1 up
    ifconfig em0 up
    ifconfig em1 up

make bridge go through ipfw

    sysctl net.link.bridge.ipfw=1

Flush the running firewall rules

    ipfw flush

create your pipes

    ipfw add pipe 1 ip from any to any via em0
    ipfw add pipe 2 ip from any to any via em1
    ipfw pipe 1 config bw 640Kbit/s queue 30 delay 64ms plr .01
    ipfw pipe 2 config bw 128Kbit/s queue 10 delay 64ms plr .01
    ipfw list

That will create one pipe (1) for outgoing traffic and another (2) for incoming traffic with a bad ADSL bandwidth allocation (bw). It will also add 64ms of latency (delay) and a small packet loss (plr)

Optional: Allocate more memory

Dummynet can end up using a lot of memory based off of how much it is being taxed. Allocate 32MB worth of memory to network mbufs.

    sysctl kern.ipc.nmbclusters=32768

Done

That should be it. You can check by pinging the gateway. You can also test by doing a speedtest. I like http://speedof.me since it uses HTML5. I have added a screenshot of my final speedtest showing this.

Questions/Comments/Priases can be sent to me at joe@unicornclouds.com