In my view point, earth is heavily exploited. The greed for money and power makes people blind. However, I wish each of us do our part to save the green earth.
Today, I would like to take you through one of the great features which seems to be ignored by the majority. As per an old statistics in Wikipedia, 67% of electricity is produced from Fossil fuels. How about saving some power in the data center ? Are we saving something beyond the financial saving ?
I just tested Power optimization and feel great on the way SCVMM is growing and proving that its mature. Though we have Dynamic Optimization and Power optimization with SCVMM 2012, I am not seeing many organizations using it actively. I would recommend you all to consider this great feature and make use of it on your next design/implementation of SCVMM. Thus we – the IT team also can make a small contribution on saving the energy and reducing the carbon foot print.
Back to the topic – What is Power Optimization? We have the infrastructure built to withstand the high load on peak business hours. On many traditional business, the load for the IT is aligned with the business hours of the active business. In such cases, we can easily predict that the chance for having a peak load in off business hours is very minimal. Why do we need to keep the full compute in the off business hours ? This thought is the base of Power optimization. With power optimization, In a defined windows – say off business windows – If the existing load on a cluster node can be distributed to other cluster nodes without affecting the load of other cluster nodes, SCVMM can turn off some nodes..
Lets try this. Power Optimization is a features which works on top of the Dynamic Optimization feature. Dynamic Optimization and Power Optimization can be configured on the Host Group Level. By default, all host groups get this setting inherited from the parent host group. So the settings which is configured on “All Hosts” host group is applied for all host groups unless the inheritance is disabled.
For configuring Power Optimization, We need the below prerequisites.
- Run As Account to be created for managing the host using Baseboard Management Controller
- Local user creation on the BMC of each host server
- Baseboard Management Controller configuration for each host on SCVMM
- Live migration should be configured for Dynamic Optimization to work
Creating a Run As Account for BMC
Its a straight process. Navigate to SCVMM -> Settings -> Security -> Run As Accounts
Please note – the user account may be case sensitive. Please check the server manual from the manufacturer for the exact details.
Now we have the Run As Account ready. The next step is to have the same run ac account created on all the Hyper V hosts.
If you have a Blade enclosure, the usual way of managing the blade is through the on board administrator, through which we can manage/login to the individual server in the blade. However, the access is getting initiated over the On Board Admin which got access to login to each server individually. This does not mean the the account you connected to the On Board Admin page can access the remote web access of individual server. Hence, we need to login to the iLO/DRAC console of each individual server and create this account on each server.
Create local user on host to access Baseboard Management Controller
The next steps walks through the procedure on an HP Blade.
Login to the OnBoard Administrator webpage.
Navigate to Enclosure Information -> Device Bays -> Server
Select Web Administration to access the iLO Web Interface of the server we selected.
From the server ILO web page, Navigate to Administration -> User Administration
On the Local User Section, Click on the “New” button which is in the right end of the screen.
Create a local user name with the same user name and password we used in SCVMM Run As Account.
Assign all privileges for this account.
Perform the same steps on all Hyper-V hosts in the cluster to create an account
Baseboard Management Controller configuration for each host on SCVMM
The next step is to configure BMC on each Hyper-V host through SCVMM. In this step, we will link each server with its remote management IP address and the corresponding run as account which has the privilege to manage it.
Navigate to SCVMM -> Fabric -> Servers -> All Hosts ->Host Group -> Server
Right Click on the Server and take the properties on the server.
Navigate to Hardware -> Advanced -> BMC Settings
Select “This physical computer is configured for out of band (OOB) management”
Select the appropriate Power Management configuration provider.
Key in the BMC IP Address.
The default port is 623. If the port got changed on the server side, please adjust it accordingly.
Select the Run As Account created for BMC.
Click on OK and verify that the job was successful on the SCVMM jobs.
Configure this setting for each server in the cluster.
We are now ready with the prerequisites.
Configuring Power Optimization
As mentioned, Power Optimization works along with Dynamic Optimization. Now we will enable Power Optimization.
Navigate to SCVMM -> Fabric -> Servers -> All hosts -> Host Group of the cluster
Right Click on the Host Group and select properties
Click on Dynamic Optimization and enable Power Optimization.
Click on Settings to configure the Power Optimization settings and the time windows.
The best candidate for Power Optimization will be identified by SCVMM and then try to see if this host can be evacuated by moving the VMs to other available nodes, at the same time keeping the destination server load with in the above configured threshold.
Schedule is the window which SCVMM can perform Power Optimization. In my case, the ideal time for power optimization is in night from 10 PM to morning 6 AM.
The time windows is based on the local server time.
In the above mentioned settings, Every day between 10 PM and 6 AM, the servers will undergo power optimization. The first node will go down if the load hosted on the first node can be moved to other nodes with out breaching the threshold of 40% Processor load and 4 GB memory. The number of servers which can be powered down is based on the total number of cluster nodes. The general rule here is the number of nodes required for satisfying node majority + 1. If its a 5 node cluster. If the cluster was created in VMM, we can use the witness disk that is automatically added to count as one additional node in this calculation.
|CLUSTER NODES||Created using SCVMM||Created outside SCVMM and later added to SCVMM|
|Maximum nodes which can be powered off||Maximum nodes which can be powered off|
Server will be in maintenance mode if picked up by Power optimization and kept shut down.
When a server which is OFF due to Power Optimization will come back ?
- If the PO schedule comes to an end. In the above example, At 6 AM every day, the servers will be powerd on by SCVMM.
- A VM in the cluster is experiencing a warning condition due to resource availability and that can only be resolved by migrating it to a powered off node