cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post

Over Committing of Storage Space

Level 10

In my previous blogs, I provided an overview of thin provisioning and discussed moving from fat to thin with thin provisioning. Now, I’d like to talk about over allocation or over committing of storage space. When you over commit storage it helps enhance application uptime. Further, it makes storage capacity management simple.

What is the over committing of storage?

When VM’s are assigned with storage more than what is actually available, it is known as over-allocation. By this mechanism, applications start their operation viewing more storage than what was actually assigned. For example, say 3 applications need 50 GB each to start operation. With over committing, all 3 can start their operation with just a total 50 GB of physical storage. The remaining 100 GB or more can be added to the storage array when the existing 50 GB’s utilization is increasing.  This way available storage arrays are utilized appropriately.

Some advantages of over committing storage are: 

  • It cuts down capital expenditures (capex): Capex will be cut down since storage space that goes unused is very minimal.
  • It allows you to dedicate more storage to VM’s than the actual available storage.
  • Flexibility of storage: there is no storage limit, more volume can be added as and when needed.
  • No trouble of forecasting storage growth: Helps you avoid the trouble of having to predict the accurate growth of volume.

Some disadvantages include:

  • Applications halt or crash: When the disk group/groups run into overcommit state (when the physical storage gets utilized 100%), applications will not have free disk to store the processed/collected data, causing the application to crash.
  • Adding free capacity can be time consuming: Manual interventions, like adding disk drives are needed to increase free capacity in the disk group/groups. And manual interventions are time consuming.
  • Chances for errors: There is a high chance for errorslike when freeing storage by deleting unwanted files or VM’s that are no longer needed, which can cause the loss of a file that had required data.
  • Rogue application:  A rogue application can completely bring down the storage as it might rapidly consume the free storage. Just imagine the rouge application sharing the same disk group as a business critical application, such as CRM, ERP, etc.

To get the best out of over committing in thin provisioning and avoid any risk, it’s important to be readily prepared. So always remember to keep a close eye on your storage. By monitoring your storage, it makes over committing and thin provisioning much easier. Furthermore, you should manage your storage or datastore by setting alerts for over allocation, so you quickly receive an SMS or email before something goes wrong. Finally, be sure to set your alerts at a decent %, so you have the necessary time to add more disk to existing storage volume.

  I hope this 3 piece blog series provided you with some helpful tips and information. If anyone has questions regarding thin provisioning, feel free to ask me in the comments section. 

14 Comments
Level 13

So basically what the airlines do with overbooking then, but hopefully with more accuracy and planning.

I can see how virtualization and storage management tools would be very useful for this.

MVP
MVP

Thank you for your insights.  Not being ingrained into the nuts and bolts of virtualization this has helped me to better understand some of the concepts.

Level 17

We have had to deal with small OS volumes. This of course once setup always gets full.

NCM Server at the moment is being adjusted, but before I can have the volumes adjusted I have to migrate my data to a temporary volume. - server has all physicals.

I had been trying to tell our server folks for months that this is going to end badly for me. It went until they got the alert that the drive was over 95% full... I had to go a night without config captures

Level 15

I have to agree that this was beneficial.  Brings to light some thoughts and wonderings that I have had on the subject.  Not my direct responsibility here at work but I am getting a better feeling for how it can be utilized.  Thanks!

Level 10

As mentioned in the blog, you should set the alerts at a decent %, so that you have enough time to add more disk or free up some space on existing storage volume.

Level 17

We are in the process of pulling most things network/server based into the same NMS. The Server/DC Folks use a different system and set their own thresholds.

  I watch my boxes closely, and request resources when needed.. long before they are needed. Sometimes the world turns a little slower than we would prefer. So As The World Turns, I keep plugging and working because after all we only have One Life To Live.

Level 15

Humor is a good thing.  Just remember "No one gets out of life alive!"

Level 17

You can always remote back in

Level 11

I wouldn't want to go any higher than 85% for the first warning, this would give you more time to respond.

Level 10

Thanks. But 85% cant be the apt % for all. This will differ based on infrastructure.

Level 17

Each environment should be monitored and assessed before determining your warning and critical thresholds.. No one wants to wake up at 3 am due to improper base lining.

Level 10

Thanks ! Great to know that this helped you.

Level 10

Absolutely right!

Level 13

I agree with you here. Even if it doesn't relate to my area, it's nice to be able to share information with teammates on what our monitoring platform can provide.