The ultimate guide to virtualization swap files!

Most operating systems natively support RAM over-extension. This is accomplished by using a temporary file on the hard drive to store RAM blocks that are not currently being actively processed so that more RAM is available for blocks that do need current processing. This file is called a swapfile or a pagefile depending on the OS. This is never a good thing, because disk is always slower than Ram speed (don't believe me? click here) But it is sometimes a necessary thing.

Managing Swapfiles and Pagefiles can be confusing in VMware or Hyper-V because there are multiple files to configure at different levels.

1. Guest VM's native OS swapfile
2. Host related swapfile
3. Hypervisor's per-guest VM swapfile

1) Guest VM's native OS swapfile

Even though you may assign 4 GB of RAM to a virtual machine, the OS running in that system may choose to create it's own pagefile on whatever virtual disks you provide according to its native behavior based on the full amount of "physical" RAM that it sees. The Hypervisor has no role in if or how a guest OS chooses to build or use a swapfile. This must be completely managed from within the Guest OS. However, remember that as a virtualization administrator you could certainly create a dedicated virtual hard drive file that is placed on a fast storage tier and then within the Guest OS ensure that this disk is used for paging. 

2) Host related swapfile

Just like the guest OS, your virtualization host has a swapfile or pagefile to support not having enough memory to perform its duties associated with being a virtualization infrastructure: (provisioning new VMs, vMotioning, etc) 

2a) VMWare System Swapfile using the Web Client: This is done via Hosts and Clusters -> Host -> Manage Tab -> Settings SubTab -> System Swap

vCenter Server Web Client configuring Host System Swap File

Edit System Swap Settings:

• Enabled - if unchecked = ALL RAM ALL THE TIME - NO SWAPFILE!
• The Datastore - Use a specific datastore
• Host Cache - use part of the host cache.
• Preferred swap file location - Use the host's preferred swap file location 

2b) Windows Server with Hyper-V: this is managed via the system control panel -> Advanced Tab -> Performance -> Settings... -> Advanced Tab -> Virtual Memory -> Change... 3) VM related swap files.

Server 2012 R2 Hyper-V server configuring Host PageFile

• If all page files are removed = ALL RAM ALL THE TIME - NO SWAPFILE!
• Don't forget to click "SET" after configuring a pagefile on a disk, otherwise nothing happened.

Hyper-V Server or Windows Server Server Core Host Swapfile

Modification is done by script or command line entries:

Add a 2 Gb Pagefile:

wmic pagefileset create name="E:\pagefile.sys"

wmic pagefileset where name="E:\\pagefile.sys" set InitialSize=2048,MaximumSize=2048

Delete a pagefile:

wmic pagefileset where name="C:\\pagefile.sys" delete

3) Hypervisor's per-guest VM Swapfile 

Modern hypervisors create a temporary swap file for hypervisor related memory management needs. This is NOT a swapfile that the Guest OS could ever see or use for its own swapfile needs! However, you may want this to be placed on an efficient disk to ensure that VMs boot as efficiently as possible. There are some key differences in how ESXi and Hyper-V use these files.

VMWare:

The host creates VMX swap files automatically, provided there is sufficient free disk space at the time a virtual machine is powered on. If it cannot be created, it prevents power on! 

This file is used because while physical memory is reserved for the system at powered on, memory for needs like the virtual machine monitor (VMM) and virtual devices can be swapped after initialization. The VMX swap feature means that a the 50MB+ for live VMX memory needs can shrink to only 10MB, freeing up memory resources for other needs. This is critical in overcommitted memory situations. 

By default, the swap file is created in the same location as the virtual machine's configuration file, but you may change the swapfile datastore datastore to another shared storage location. Moving the swapfile to a local datastore may improve local performance, but it may also slow vMotions later because pages swapped to a local swap file on the source host must be transferred across the network to the destination host

Hyper-V:
Smart Paging in Hyper-V is used only when the following is true:

• The VM is being restarted (directly or via a host restart)
• The hypervisor discovers that there is no available RAM
• No memory can currently be reclaimed any other VMs running on the host

At that time the Smart Paging file is used by the VM as memory to complete startup. Within 10 minutes, that memory mapped to the Smart Paging will need to be provisioned into RAM and the the Smart Paging file will be deleted.  The Smart Paging feature is ONLY to provide reliable restarts (not cold boot or running out of RAM later) of VMs

3a) How to configure the vSphere web client to configure VM support swapfile: Hosts & Clusters -> Host -> Manage Tab -> Settings SubTab -> Virtual Machines -> Swap File Location

vSphere Web client configuring Default Swap File Location

vSphere Desktop Client to configure VM support swapfile: Select Host -> Configuration Tab -> Software Group - Virtual Machine Swapfile Location

This setting can also be overridden on a per-VM basis if needed in the VM Swapfile Location property

3b) Hyper-V SmartPage: Select a VM -> Right Click (Or Actions) - Settings -> Smart Paging File Location

The file location is determined during installation, then reconfigured on a VM by VM basis in their management related properties:





I hope that you feel a little more solid on the 3 types of swapfiles that you will run into when managing virtualization!

Keep it virtual!

Compare Seek Times Between RAM, HDD, and SSD Hard Drives

How much faster is RAM than a Hard Drive anyway?

Let's look at some specs, do some math, and make an analogy to help us relate!
First, let's compare the retrieval time from different media types:

• RAM: 3 nanoseconds
• Mobile Disk = 12 milliseconds = 12,000,000 ns
• Standard Disk - 9 milliseconds = 9,000,000 ns
• Server Class HDD = 3 milliseconds = 3,000,000 ns
• Server Class SSD = 0.1 milliseconds = 100,000 ns

When we compare these media, we see that RAM is at least a 1 million (and likely 4 million) times faster than HDD and 33,333 times faster than SSD. But how do you even conceive what those kinds of numbers mean?

Let's play a game where nanoseconds become regular seconds just to get an idea of comparative speeds.

Imagine you are at a restaurant, and you ask the server for some ketchup (or catsup if you prefer) for your meal. The server looks at the adjacent empty table, grabs the available bottle of ketchup, and hands it to you at RAM speed: total time = 3 seconds.

 

 

Now if your server was operating like a standard High Speed Server HDD, it would take them... 3,000,000 seconds = 34.7 days = over a MONTH to get you the ketchup!

Heaven forbid your sever is using a Standard HDD because that would be 9,000,000 seconds = 104 days = 3.4 months!

And if your server is grabbing your "ketchup" from a Mobile HDD, it would be even worse: 12,000,000 seconds = 139 days = 4.6 months!

Even optimized with a Server class SSD your server is probably going to be slower than you would want to wait: 100,000 seconds = 1.15 days.

Hopefully the lesson is not lost on all of you:
Cache is KING!

Invest in your RAM - and look for options in your software to optimizes RAM and cache usage!

That is all.