Memory Management

kbjijo
kbjijo used Ask the Experts™
on
Hi
I wants to know when & how the  esx host manage teh meory
Esx host have 98 GB RAM and hosted 10 VM ( each VM have 24GB RAM)
I can see the Host Mem utilisation is 92GB. (Shared by all VM and directly consuming from Physical RAM)
At VM level, there is no value for Ballooning & swaping.

1 I wants to know why the Ballooning and swapping is not happening ?
2 If VM are occupied more than 90 % in Physical memory, whether i can build anotehr VM with Large memory in the same host ?
3 At above said condition(92 Gb utilised), if all VM are requesting for their granted memory, then how the esx host will manage ( This situation, ESX already triggered Critical Mem usage Alarm)


Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
Have you got VMware Tools installed in the Virtual Machines, because it has the memory balloon driver?

If all machines request their memory, the ESX host will start to swap to disk, causing performance issues.

Have a read of this Yellow Brick article

Which metric to use for monitoring memory?

http://www.yellow-bricks.com/2011/04/29/which-metric-to-use-for-monitoring-memory/

http://justincockrell.com/?p=125

Commented:
esx 4 does memory slices better than 3 so what version esx

Author

Commented:
Thanks for the reply,

All VM have VMtoolinstalled, and our version is ESXi 4.1
One Doubt:please correct me if i am wrong.
When enough resource is available in Physical RAM, then ballooning & swapp is not rquiired,
I will clear my question
eg:
ESX : 8 GB RAM ,
VM1 : 6GB (Current Host Mem usage is 5GB)
VM2 : 6GB (Current Host Mem usage is 2GB)
Here ESX have enough resource to allocate the memory from RAM. But it generating alrm 90%.
Next time, VM2 required another 2GB, then how the ESX will manage the overcommitt? what is the esx host warning level ?




If the total amount of RAM required by the VMs exceeds the amount of physical RAM that ESX has, then the VMs will begin swapping to disk based on a first-come, first-serve priority if no resource shares or resource pools are configured.

An ESX host can manage overcommitment of guest Virtual Machine memory through a variety of technologies, but eventually a point is reached where the overcommitment is too much.

ESX/ESXi 4.1 follow this process:

1. Lots of available physical RAM - assign memory to VMs as needed.

2. Greater than 95% utilization of physical RAM - break large memory pages (2 MB) into small memory pages (4K).  Implement Transparent Page Sharing (effectively de-duplication of memory).  This will lower the overall utilization shown for RAM, dropping it 5-35% depending on VM usage.

3. When utilization again reaches around 95% or greater, memory compression kicks in and further reduces the % of physical memory in use.

4. When utilization again reaches around 95% or greater, VM Ballooning kicks in to reclaim unused memory from any VMs with VMware Tools installed.  If unused memory is found, the % of physical memory in use again goes down.

5. Lastly, when no other options are available, VMs swap to disk.


Each of the steps above causes a performance hit.  With low-use VMs or Desktop VMs, this performance hit is likely not to be noticed until the VM reaches a point of swapping to disk.   With production and high I/O VMs, performance may be impacted at an earlier stage.

This leads to the "it depends" answer when one asks how over-subscribed memory on an ESX/ESXi host can be.  It depends on the SLAs and needs of the VMs running on the host.  Testing in your specific environment will help determine the ratio at which over-subscription of RAM can set at.
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017
Commented:
you've currently overcommitted the RAM, you've allocated in the virtual machines.

Both VMs, have 6GB, total 12GB,

ESX host has only 8GB.

So therefore, if all virtual machines needed there full access to 6GB each (12GB), ESX server would start swapping it's memory to disk, using the swap file and performance (ESX host) would be very poor, causing poor performance for the virtual machines.

either ensure the Virtual machines do not use 6GB, re-tune them, and reduce the memory allocate to the, do they reallt need 6GB each, or purchase more memory for the host.

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial