Is Memory over commits kill the esx server if the usages more than a physical memory..?

[kathirkk23]

Hi,

I am running esx 3.5 on a 4GB DELL physical server and more than 5 VM's running under that .

Each VM's I assigned 2 GB RAM that means 5X2 =10 but physical server Memory is only 2 GB , even though I assigned bcos vmware supports

memory over commits, My question is if all the 5 VM's really wants to 2 GB means how my H/W server will support...?

I observed in lab that if all tries to utilize full then my esx server keep on restaring...

Is any idea/suggestion to avoid this on ESX 3.5

Kathiravan.K

[AWo]

Welcome to the forums!

I am running esx 3.5 on a 4GB DELL physical server and more than 5 VM's running under that .

Each VM's I assigned 2 GB RAM that means 5X2 =10 but physical server Memory is only 2 GB , even though I assigned bcos vmware supports

4 or 2 GB?

However, you can overcommit memory, but that's one thing you should avoid. If the guests really request more memory than what is physically available, ESX starts to swap (as other OS's do when more memory is needed). But this mechanism is more to keep things alive, not fast. You will suffer from a performance degradation.

[TomHowarth]

Welcome to the forums

Firstly that amount of over-commit is not a good thing. if your guest really require that amount of memory then you really need a lot more memory in your Host machine.

Let me explain

Firstly ESX requires memory to run, upto 800MB depending upon the amount of memory you assigned to your service console. this therefore leave 3.2GB at the worst case for running guests, or 320B per VM. you are running at 2GB per Guest an over-commitment of 1680MB per Guest. it is no wonder you are having issues. I would look at your guests and lower their memory to 512MB and then monitor, you can always increase the memory of certain Guests if needed.

Memoy overcommit is best used where you have a very good handle of the actual memory requiement of your guests, ie you have monitored them over a period of time and noted that although machine X has 4GB of RAM, and had been monitored to utilise only 2GB, (this means you can over commit by 2GB with little expectation of Swapping to disk occuring)

do your Guests have VMtools installed, if performance is not too much of an issue (ie it is a dev/test environment) you could introduct the Memory balloning driver into the mix this function of VMtools will move little moved memory space out to disk. (similar in concept to a page file in windows) it does however have quite a significant effect on performance.

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

Tom Howarth

VMware Communities User Moderator

[kathirkk23]

Thanks Tom Howarth, AWO for your valid point.

But my case is different,our ESX server is automatically restarting when all the VMs need or utilizing more memory .hence I allocated 2 GB to 4 VMs where my physical server memory itself only 4GB.

Regards

Kathiravan

[weinstein5]

also one thing Tom did not mention is the Transparent Page Sharing - this is the vmkernel's ability to recognize identical pages being stored by the VMs and will only store them one time in Copy on Write memory - this saves a ton of memory of vms running identical operating systems and similar applications - this is one of the tools that allows for overcommittment -

If you find this or any other answer useful please consider awarding points by marking the answer helpful or correct -

[kathirkk23]

Thanks Toms,

See in my scenorio all of my 4 VMs not using 2 GB memory all the times, say for ex. my epayroll VM's only need more momory on end of the month.

and attendance application VM need more memory on only 1st week of the month thats y I have assigned each VM's 2 GB.

Regards

Kathir