Wednesday, September 15, 2010

Presentation

Almost finished the presentation at this stage, just a matter of adding in the final experiments!
60slides plus :(

Thursday, August 26, 2010



Finish testing..



I've also structured the layout of my paper...

Tuesday, August 17, 2010

Tests against each Experiment


Within each experiment there are 5 tests which are ran...

Test 1 - Geekbench - performs a number of general performance measurements against a system.

Test 2a - Ramspeed test - Performs 4 operations : Copy(A=B), Scale(A=B*m), Add(A=B+C) and Triad(A=B+C*m). using INT
Test 2b - Ramspeed test - Performs 4 operations : Copy(A=B), Scale(A=B*m), Add(A=B+C) and Triad(A=B+C*m). using FP

Test 3 - systester - 5 runs are made against calculating a 512K Pi value - these runs are totaled & timed - CPU based test

Test 4a - Disk Subsystem - Performs 10 random 32mb reads against the disk - noting performance information

Test 4b -Disk Subsystem - Performs 10 random 32mb writes against the disk

Test 5a - Network connectivity to the network gateway - physical device on the network
Test 5b - Network connectivity to the Host - via virtual networking
Test 5c - Network connectivity to another VM within the same host

Test6a- Forkbomb executed on 1 VM within host
Test6b- Forkbomb executed on 2 VM's within host
Test6c- Forkbomb executed on 3VM's within host

All tests are carried out from a Linux VM within the host.. Also 2 control experiments... 1 from within a VM with no hostile events taking place...

I've finally revised all my test scripts..

with 2 seperate tests for CPU/Memory/DiskIO/NetworkIO

I've completed testing on Virtualbox & Workstation

fully configured XEN and ESXi & Hyper-V

Hypervisors & Experiments

Hypervisor1- VirtualBox running on Windows 7 64bit - (Sata disk 1)
Hypervisor2 - Workstation running on Windows7 64bit - (Sata disk 1)
Hypervisor3 - Hyper-V within Win2008 R2 - (Sata disk 1)
Hypervisor4 - XENServer - (Sata disk 2)
Hypervisor5- ESXi - (Sata disk 3)

Experiment 1 - Crashme - sending non-sensical data to a VM (WinXP least secure) with the intent of causing a full crash of the VM itself, if not a full crash the vCPU will have to deal with non-sensical instructions

Experiment 2 - Fuzz Testing - sending random messages via specific programs causing CPU & Mem overloading of a VM, this can cause some application to crash, however my intent is not to test FUZZ, however see it's effect on other VM's within the same Host.

Experiment 3 - Fork Bombs... - By initializing a fork bomb - (causing a process to create unlimited child processes ) can over load the memory within a VM, causing it to become unstable or actually hang/crash - Divided into 3 sub experiments in which 3 out of 4 running VM's are hit with Forkbombs

Experiment 4 - DoS attack - Attack using 2 zombie machines (not located on the host) to attack using a TCP syn storm a target VM within the hosted environment, This server should become unstable, let alone the flooding of the v network cards / physical network card.


Hypervisors dropped:

Parallels running on Windows 7 64bit (Type 2 Hypervisor)
Reason: Unable to run experiment 4 on target VM
Issue: Parallels only uses NAT for virtual networking from Virtual Machines via host NIC through to the physical network, therefore the VM's IP address is a private IP inside the Host on the Virtual Switch.

In order for Experiment 4 to work Bridged networking must be in place, which using dynamic IP's via a DHCP service or Static IP's each VM has it's own MAC & IP address visible through the physical NIC on the physical network.


KVM runnning on ubuntu 10.4
Reason: Same as above, unable to configure virtual switch for bridging & also unable to grab/ungrab VM's desktop for MS machines using the AQEMU management console within KVM.

As a result via both problems I was unable to test and even configure all test VM's on this platform due to AQEMU freezing up.