Question Why ever does Parallels 8 recommend max 4 GB RAM for a VM?

[ovalsquare]

As per my question up at http://apple.stackexchange.com/questions/65640/why-ever-does-parallels-8-recommend-max-4-gb-ram-for-a-vm, I'm a VMware Fusion user and am trying out Parallels since VMware has some quirks that are annoying on Mountain Lion on a retina MacBook Pro (16 GB RAM).

I'm wondering about the baffling Parallels recommendation (when configuring a VM) that there is a maximum cap of 4 GB of RAM for a VM with Parallels (indeed, it recommends 1.5 GB for Windows 8). (See forum thread at http://forum.parallels.com/showthread.php?t=114040). Granting anything more would, in fact, slow down both the host and client, but I have no problem assigning 8 GB of RAM to a VM in Fusion; before really committing to Parallels, I'd like to know the rationale behind such a recommendation (and no, the Parallels forums don't shine satisfactory light on the subject, at least from what I've been able to dig up).

If you're running in an x64 environment for both host and client then there shouldn't be any acceptable reason for needing to cap the RAM. Apparently, Parallels support states "because Windows is working in a virtual environment, that it only needs max 1-3 GB." My development machine and environment is not a toy, so such an answer seems borderline patronizing and makes me think they're simply covering up for a significant Parallels limitation.

 

[luciferano]

I don't know exactly why they make that recommendation, but that you say it would slow things down and that VMs usually aren't used in such a way that more than 1-2GB is necessary for Windows Vista through Windows 8 and their server versions seem like rather convincing reasons. They might be covering up for something, but it's not like there's no other plausible explanation.

EDIT: If I had to guess what they're covering up, assuming that they are, it might be that the 4GB limit is somehow related to 32 bit programming. I know that some other virtual machine programs have no such limitation, so this might be the issue.

 

[Pinhedd]

As per my question up at http://apple.stackexchange.com/questions/65640/why-ever-does-parallels-8-recommend-max-4-gb-ram-for-a-vm, I'm a VMware Fusion user and am trying out Parallels since VMware has some quirks that are annoying on Mountain Lion on a retina MacBook Pro (16 GB RAM).

I'm wondering about the baffling Parallels recommendation (when configuring a VM) that there is a maximum cap of 4 GB of RAM for a VM with Parallels (indeed, it recommends 1.5 GB for Windows 8). (See forum thread at http://forum.parallels.com/showthread.php?t=114040). Granting anything more would, in fact, slow down both the host and client, but I have no problem assigning 8 GB of RAM to a VM in Fusion; before really committing to Parallels, I'd like to know the rationale behind such a recommendation (and no, the Parallels forums don't shine satisfactory light on the subject, at least from what I've been able to dig up).

If you're running in an x64 environment for both host and client then there shouldn't be any acceptable reason for needing to cap the RAM. Apparently, Parallels support states "because Windows is working in a virtual environment, that it only needs max 1-3 GB." My development machine and environment is not a toy, so such an answer seems borderline patronizing and makes me think they're simply covering up for a significant Parallels limitation.

They recommend 4GiB because there's a significant amount of storage overhead required to store a guest on the host. I'm not sure how Parallels works but VMWare Workstation / VMWare Fusion write the entirety of the virtual memory to the hard disk when the guest is suspended and this same file is locked when the guest is running. Since most operating systems use non-committed memory to cache disk data this also gets written to the disk. Snapshots, etc... also maintain disk memory. Thus, the implications of additional memory to a guest are more than simply denying the host the use of that same memory, it can cause disk IO to increase and this can have a negative impact on the guest performance.

 

[Ijack]

Like the others, I don't know the answer to your question. But I do wonder why people pay for VM software when the excellent VirtualBox does the job for free.

 

[luciferano]

Like the others, I don't know the answer to your question. But I do wonder why people pay for VM software when the excellent VirtualBox does the job for free.

VMWare and QEMU too

 

[Ijack]

I dont believe that VMWare player is available for OS X. Qemu is good, particularly if you want to emulate Other than x86, but doesn't have the features that VirtualBox has.

 

[luciferano]

I dont believe that VMWare player is available for OS X. Qemu is good, particularly if you want to emulate Other than x86, but doesn't have the features that VirtualBox has.

VMWare is available on OSX. It's under a different name, but it's there. I think that it's called VMWare Fusion or something like that. I've used it on friends' computers.

However, Virtual Box is my preference too.

 

[Ijack]

Well yes. But that's the expensive software that the OP was talking about.

 

[Pinhedd]

Like the others, I don't know the answer to your question. But I do wonder why people pay for VM software when the excellent VirtualBox does the job for free.

Virtualbox may be free but it's abhorrently buggy to a degree which is not acceptable in a production environment. There are still critical IO issues with the shared drives that have been present for ages. The bug tracker is a shining example of why free is not always better

 

[luciferano]

Virtualbox may be free but it's abhorrently buggy to a degree which is not acceptable in a production environment. There are still critical IO issues with the shared drives that have been present for ages. The bug tracker is a shining example of why free is not always better

I'm not saying that my experience is indicative of everyone else's, but I've never had a bug problem with Virtual Box. I've also never had one with VMWare, for what that's worth.

 

[Pinhedd]

I'm not saying that my experience is indicative of everyone else's, but I've never had a bug problem with Virtual Box. I've also never had one with VMWare, for what that's worth.

I have. The shared folders in virtualbox are a nightmare. They slow down exponentially as files are added and when small files (usully 4KiB or less) are written they get out of sync.

For example,

use a script which creates a directory, then creates a few files in that directory, then deletes the files, and then deletes the directory. Do this with a moderate number of files and folders in rapid succession. The script may fail due to one or more IO errors.

Similarly, adding files or folders to a shared folder on the host side will not reflect properly on the guest side to services which are listening for file changes (such as a static file server).

This problem is still present as of VirtualBox version 4.2.0 and has been present since at least version 4.0.8

Using NFS avoids this issue but may require root privileges on the host.

 

[luciferano]

I have. The shared folders in virtualbox are a nightmare. They slow down exponentially as files are added and when small files (usully 4KiB or less) are written they get out of sync.

For example,

use a script which creates a directory, then creates a few files in that directory, then deletes the files, and then deletes the directory. Do this with a moderate number of files and folders in rapid succession. The script may fail due to one or more IO errors.

Similarly, adding files or folders to a shared folder on the host side will not reflect properly on the guest side to services which are listening for file changes (such as a static file server).

This problem is still present as of VirtualBox version 4.2.0 and has been present since at least version 4.0.8

Using NFS avoids this issue but may require root privileges on the host.

I wouldn't know about that. I just use regular network shares for my virtual machines.

 

[Ijack]

I've had more problems (on Windows) with VMWare than with Virtual Box - mainly networking problems. I've only used VirtualBox on OS X so I don't know if the same problems exist there. But I can appreciate that if VirtualBox bugs bite youthen it is not for you.

 

[ovalsquare]

Thanks to all for replies. Thought I'd get notifications by default, but guess not as didn't get any notifications as to replies.

As I stated originally, the 4GB limit could have to do with x64 environments (64-bit), but if so, no other mainstream virtual solution has this limit, so I would say that this is an unnecessary and unacceptable limitation.

They recommend 4GiB because there's a significant amount of storage overhead required to store a guest on the host. I'm not sure how Parallels works but VMWare Workstation / VMWare Fusion write the entirety of the virtual memory to the hard disk when the guest is suspended and this same file is locked when the guest is running. Since most operating systems use non-committed memory to cache disk data this also gets written to the disk. Snapshots, etc... also maintain disk memory. Thus, the implications of additional memory to a guest are more than simply denying the host the use of that same memory, it can cause disk IO to increase and this can have a negative impact on the guest performance.

As to increased disk IO, you'd think providing additional RAM would ultimately reduce the disk IO: even if other aspects of it would marginally increase it, just the guest itself would not have to page any memory to disk cache. Heck, that's the biggest reason for performance increase with more RAM - less disk IO. Which by the way neither VMware Fusion nor VirtualBox have this limitation (and I've had no problems running 8GB for the guest on the former).

Therefore, I'd argue that that disk IO is mostly only apparent when suspending or resuming as any additional RAM would more than make up for any additional disk churn. A more than acceptable trade-off.

So we're back to a Parallels limitation that isn't satisfactorily explained and could be not even indeed necessary if your guest is x64 and you've got plenty of RAM to go around.

 

[ovalsquare]

Like the others, I don't know the answer to your question. But I do wonder why people pay for VM software when the excellent VirtualBox does the job for free.

Have a co-worker who swears by VirtualBox, but the share performance and lack of slick integration is the deal breaker for me. Both Parallels and VMware Fusion make the shares as seamless as I think would be possible which is an absolute requirement for me. Thus it came down to paying for Parallels or Fusion. With upgrade pricing on the former and $49 for the latter, it isn't horrible, unlike VMware Workstation on Windows which is >$150 even with upgrade pricing; albeit is considerably more powerful and the latest is great for managing vSphere VMs from the local machine.

 

[ovalsquare]

I don't know exactly why they make that recommendation, but that you say it would slow things down and that VMs usually aren't used in such a way that more than 1-2GB is necessary for Windows Vista through Windows 8 and their server versions

Beg to differ on that 1-2GB bit ;-) Yes, runs fine. But if it's not a toy and you're running a development environment, the more RAM the better. I'd be running 8GB on my Windows 8 dev environment and I would experiment with running it up higher depending on whether or not the remaining 8GB was good on the host.

 

[Pinhedd]

Thanks to all for replies. Thought I'd get notifications by default, but guess not as didn't get any notifications as to replies.

As I stated originally, the 4GB limit could have to do with x64 environments (64-bit), but if so, no other mainstream virtual solution has this limit, so I would say that this is an unnecessary and unacceptable limitation.



As to increased disk IO, you'd think providing additional RAM would ultimately reduce the disk IO: even if other aspects of it would marginally increase it, just the guest itself would not have to page any memory to disk cache. Heck, that's the biggest reason for performance increase with more RAM - less disk IO. Which by the way neither VMware Fusion nor VirtualBox have this limitation (and I've had no problems running 8GB for the guest on the former).

Therefore, I'd argue that that disk IO is mostly only apparent when suspending or resuming as any additional RAM would more than make up for any additional disk churn. A more than acceptable trade-off.

So we're back to a Parallels limitation that isn't satisfactorily explained and could be not even indeed necessary if your guest is x64 and you've got plenty of RAM to go around.

I'm not sure why they have a 4GB cap. It is possible to run a 32 bit guest on a 64 bit host with more than 4GiB of memory if PAE is virtualized. It is also possible to run a 64 bit guest on a 32 bit host provided that the host supports PAE.

 

[ovalsquare]

Well, look at that. Latest update released today (October 26, 2012) now allows "assigning up to 16 GB of memory to a single virtual machine."

See http://kb.parallels.com/115024

You can now assign up to 16 GB of memory (RAM) to a single virtual machine.
Support for Windows Server 2012.
New, more intuitive design to optimize Windows for working on Retina displays.
Easily check the reclaiming disk space progress with a new progress bar.

However, I've installed and checked, and it still says "For best performance, set available memory to be within the recommended range of 512 MB to 4096 MB. Allocating memory outside of this range can slow down both the guest OS and your Mac."

The mystery continues...

 

[somedude82]

As per my question up at http://apple.stackexchange.com/questions/65640/why-ever-does-parallels-8-recommend-max-4-gb-ram-for-a-vm, I'm a VMware Fusion user and am trying out Parallels since VMware has some quirks that are annoying on Mountain Lion on a retina MacBook Pro (16 GB RAM).

I'm wondering about the baffling Parallels recommendation (when configuring a VM) that there is a maximum cap of 4 GB of RAM for a VM with Parallels (indeed, it recommends 1.5 GB for Windows 8). (See forum thread at http://forum.parallels.com/showthread.php?t=114040). Granting anything more would, in fact, slow down both the host and client, but I have no problem assigning 8 GB of RAM to a VM in Fusion; before really committing to Parallels, I'd like to know the rationale behind such a recommendation (and no, the Parallels forums don't shine satisfactory light on the subject, at least from what I've been able to dig up).

If you're running in an x64 environment for both host and client then there shouldn't be any acceptable reason for needing to cap the RAM. Apparently, Parallels support states "because Windows is working in a virtual environment, that it only needs max 1-3 GB." My development machine and environment is not a toy, so such an answer seems borderline patronizing and makes me think they're simply covering up for a significant Parallels limitation.

So did you set it to 8GB for your VM and did it perform as intended?

 

[ovalsquare]

So did you set it to 8GB for your VM and did it perform as intended?

Currently running with just 4GB for each VM and it works fine for existing projects. Haven't bothered with more atm because there's been no need (and the suspend images really add up for my poor 256GB drive the more RAM is allocated - really should have splurged on the ridiculously overpriced 512GB option). When working on a project that requires more we'll find out ;-)

 

[ovalsquare]

Long afterwards, figured I'd update this post. Parallels 9 now has sensible recommendations.