ece1508 - Energy-efficient Telco Cloud - HW8 - Solved

Network Softwarization: Technologies and Enablers

 Note: This a team work assignment. Each team (3 or 4 students) will work on a dedicated blade server. The students register their names (or emails) at following URL: 

https://docs.google.com/spreadsheets/d/1WXikPNqUSScsOtX4oPX3Tup8LCUUA-

DtYgxaXHld_D8 

1. Connecting to an Ericsson blade server (EBS) 

OpenVPN is required to connect to the experimental environment. Download and install OpenVPN client from:  https://openvpn.net/index.php/open-source/downloads.html 

Each team will receive following information:

•       OpenVPN key files (sde-xxx.ovpn) for each member to access the system

•       Private key file “create-netsoft.ppk” (or “create-netsoft.pem” in case you use Linux) to open a SSH connection to an EBS  

•       IP and credential information of the EBS and shifted OpenStack platform.

In order to open a VPN connection, copy the provided keyfile to OpenVPN client config folder. In the Windows environment, it is:  

C:\Program Files\OpenVPN\config\

Launch OpenVPN to connect to the VPN network:

  

Launch a SSH client (e.g., putty) to connect to the Ericsson blade server. The putty software can be downloaded from:

https://www.chiark.greenend.org.uk/~sgtatham/putty/latest.html 

  

Import the SSH private key of the blade (“create-netsoft.ppk” file), as follows:

  

And then click “Open” to connect to the blade with provided EBS credential information:

  

 

When you are in the blade, create a folder to save your work.

#mkdir your_group_name

2.       Connecting to OpenStack 

To connect to OpenStack on your blade server, simply open a browser, and point to: http://your_blade_ip with the provided credential information. In the previous example, it should be:  http://172.29.2.11/ 

Refer to the previous lab for the manual of OpenStack.

3.       Helper commands 

The following commands may be required to achieve your assignement.

Command 
Utility 
#sudo apt-get install cpufrequtils
Install the cpufrequtils package to change CPU frequency if it is not yet available in your blade
#cpufreq-info
Determine the current frequency of each CPU by the command
#sudo apt-get install sysstat
Install the systat package to collect CPU utilization
#mpstat -P ALL
Determine the load of the CPUs in the blade
#cat

/sys/devices/system/cpu/cpuX/cpufreq

/scaling_available_frequencies
Determine the available frequencies of the CPU X (0 ≤ X ≤ 11)
#sudo sh -c "echo -n "userspace" >

/sys/devices/system/cpu/cpuX/cpufreq

/scaling_governor"

#sudo sh -c "echo -n YYYY >

/sys/devices/system/cpu/cpuX/cpufreq

/scaling_setspeed"
Changing the frequency of the CPU X

(0 ≤ X ≤ 11)

YYYY is an available frequency (see the command above)
#sudo sh -c "echo -n 0 >

/sys/devices/system/cpu/cpuX/online"
Disable the CPU X (1 ≤ X ≤ 11)

Please do not disable the CPU 0
#sudo sh -c "echo -n 1 >

/sys/devices/system/cpu/cpuX/online"
Activate the CPU X (1 ≤ X ≤ 11)
#sudo eri-ipmitool rns 0 4 5 68 69
Collect data from temperature, voltage, and current sensors
# sudo apt-get install python-heatclient
Install the command-line OpenStack Heat client
 

4.       Assignment (50 points) 

You are required to implement five small scripts as follows. The scripts would be written in Python or in shell (e.g., bash) languages. You may also combine the scripts into a single file if it is necessary.

1.       (15 points) Write a script CollectE to collect data from the blade server in every second. Data can be saved into a .csv file, including the following fields (fields are separated by “;”):

•       Timestamp

•       Temperature

•       Power1

•       Power2

•       Total power

•       CPU0 utilisation (%)     CPU0 frequency

•       CPU1 utilisation (%)

•       CPU1 frequency  

•       …

•       CPU11 utilisation (%)

•       CPU11 frequency

The script should be run in the background to collect data in real-time.

Recall: Power1 = Voltage1 x Current1; Power2= Voltage2 x Current2; Total power = Power1 + Power2.

 

2.       (5 points) Write a script CreateE to create a random number N of VMs (you may freely choose the bounds of N, for example: 1 ≤ N ≤ 8). All VMs are created from the same flavour (for example: tempest2, 1 vCPU, 512M RAM, 1G root disk). It is recommended to use OpenStack HEAT to create the VMs.

 

3.       (5 points) Write a script ArrivalE that launches the script in (2) in a random interval of time λ (you may freely choose the bounds of λ, for example: 1min ≤ λ ≤ 5min). Run this script within a time T (T = 30 minutes).

 

4.       (5 points) Write a script ServE that destroys the VMs created in (3) in after random interval of time μ (you may freely choose the bounds of μ, for example: 1min ≤ μ ≤ 5min). Because of the randomness in (3) and (4), there could be multiple VMs coexisting in the system in the same time. The script in (4) will end when ArrivalE stopped and all the VMs have been destroyed.

 

5.       (10 points) Write a script AdaptE that adapts the CPU frequency and active/disabled state to save energy consumption. The script may check the CPU utilization, and change it to an optimal frequency according to the requests of creating VMs.

 

6.       (10 points) You are required to discuss the following items in a report. Please use graphical charts to illustrate your statement.

a.       What is the average creation time (λ)?

b.       What is the average running time of a VM (μ)? In average, how many VMs are there in the system?

c.       In average, how much energy does a VM consume? How much energy does it consume for each phase: creation, running, destruction? What is the average temperature in each phase?

d.       How does your energy adaptation strategy work?

e.       What is your saving energy (compared to a case in which no adaptation is implemented)?

f.        Propose a model of energy consumption and temperature according to the number of VMs for the: i) regular case (no adaptation), and ii) your proposed adaptation strategy.