VMware VCM 5.3 User Manual Page 30

  • Download
  • Add to my manuals
  • Print
  • Page
    / 31
  • Table of contents
  • BOOKMARKS
  • Rated. / 5. Based on customer reviews
Page view 29
Sizing Memory for Virtual Machines
30
Deploying Microsoft Windows 7 Virtual Desktops with VMware View Applied Best Practices
Introduction
The amount of RAM allocated to a virtual machine has a direct effect on a number of factors that
impact both capacity and performance requirements for a virtual machine. This appendix briefly
explains how to allocate RAM to Windows 7 virtual machines. For the sake of simplicity,
memory overhead of the ESX host will not be included in any calculations.
This appendix also assumes that users have a good understanding of ESX memory management
technologies such as memory over-commitment, transparent page sharing (TPS) and vswap
usage. More information can be found in the vSphere Resource Management Guide found at
www.vmware.com.
As an advanced configuration consideration, if you do not over-commit memory, then it is
possible to disable TPS which will free up CPU cycles which the virtual machines can use. TPS
can be disabled by setting
Mem.ShareScanGhz to zero under the advanced settings for an ESX
host. This can provide a 5 percent performance boost to the virtual machines on the host.
Virtual Machine Active Memory (Working Set)
The amount of RAM that is actively used by a virtual machine is referred to as the active
working set. This can be seen with the %ACTV counter in esxtop or the “Active” memory
counter in vCenter under the
Resource Allocation tab. On a virtual machine with 2 GB of
RAM, if the max working set size is 1GB (%ACTV == 50%) then the virtual machine is using 50
percent of its allocated RAM. Assuming other virtual machines exhibit similar behavior, a 2:1
over-commit ratio could be used without causing excessive swapping by the ESX host.
On the ESX server, if the active working set for all virtual machines is less than the total
available host memory, then all the virtual machines run at full speed because each virtual
machine is able to address all the RAM it requires without the hypervisor ballooning or swapping
virtual machine memory pages to disk.
Conversely, when the sum of all active working sets on the host exceeds the amount of available
RAM on the ESX host, the hypervisor will be forced to swap pages from the virtual machine
memory to vswap. The hypervisor has no knowledge of which pages are in the active working set
of the virtual machine and will swap the pages to meet the memory demands placed on the ESX
host. This will lead to very poor performance of the virtual machine and should be avoided at all
times.
Refer to Table 1 for an example in which a host with 32 GB of RAM is hosting virtual machines
each co
nf
igured for 2 GB of RAM. Each virtual machine has on average 50 percent active
memory.
Table 1 Memory over-commit
VM Count Active Memory in host Comments
16
16 * 2 GB * 50% = 16 GB
Without memory over-commit, only 50% of the host’s
memory is actively in use.
32
32 * 2 GB * 50% = 32GB
Memory is over-committed by 200% but active memory
is equal to host memory. Virtual machines will run at
full speed until usage exceeds 100% of host memory.
48
48 * 2 GB * 50% = 48 GB,
limited to
32 GB
by host
These virtual machines want 48 GB of RAM but are
limited to the 32 GB that is installed on the host. ESX
must swap to allow these machines to run
and performance will be degraded.
When used appropriately, the memory over-commit feature allows virtual infrastructure
administrators to drive the ESX hosts to high memory utilization without degrading performance
which is not possible with most physical systems.
Page view 29
1 2 ... 25 26 27 28 29 30 31

Comments to this Manuals

No comments