PD 10.1.1 Causing Host OS To Hang W/ Intensive Disk I/O Procedures?

[PhnordP]

I'm on a MacBook Pro w/ 16GB of RAM, 2.6GHz Intel Core i7 eight-core processor running OSX 10.9.5. I'm using Parallels is 10.1.1 (28614), and have found this issue to happen when running either Windows 7 & Windows 8.1 as guest OS.

I first noticed this problem when I was installing Windows 7: as it installed, my *whole* system would slow to a crawl. I'm not just talking about spinning pinwheels; I mean it would slow to the point that hardware was no longer operational - the screen would not update, my cursor would get stuck in place on the screen for periods of a good 1-2 minutes before being moveable for 30 seconds and then hanging again, etc. etc.

This happened during the installation, which ended up taking a very, very long time as a result. However, once installed it stopped happening... until I tried to update the OS. Then I had the same issue occur.

Same thing happens in Windows 8.1.

It also happened during the installation of MS Office into either of the above.

It also sometimes happens while Suspending my Windows 8.1 VM (haven't noticed this on the 7 installation, but I've been using that far less than the other).

I thought maybe it was a performance issue - although that wouldn't make sense as I have an exceedingly souped-up MAC - so I cut it down to devoting 1/2 of my 8 cores & 1/2 of my 16GB of RAM only to the VM (4 cores & 8 GB, respectively); still had the same issues, although perhaps less frequently? Hard to say...

This has never happened when using my Mac in any other fashion; only in Parallels.

The common denominator *seems* to be heavy disk-reading-and-writing processes, although that's hard to say for certain. I have a 1 TB SSD drive installed, with a good 100 GB free...

Any thoughts? Any ideas of how to resolve it, or what log files to post / reference / scour for details? I'm at my wits end here...

 

[AlekseyM]

Does it happen right after boot or when you are installing/loading programs? What antivirus software do you have installed? Generally AV checks every file that is open/saved and loaded.

Can you also check if parallels is the one eating resources. Open up activity monitor (use spotlight to find). Sort by CPU and run the VM and see if it spikes.

 

[PhnordP]

AleksyM - I explained in great detail when it happens in my initial post; to repeat: it happens most when I am installing an OS, a lot when I am updating the OS, some when I am installing other software, and occasionally while suspending a VM.

I have all active AV disabled as I am only using my Guest OS with safe, reliable software and would rather not see a performance hit from it.

As it's difficult for me to replicate updating the OS again, as it is currently up to date, I observed the CPU usage while suspending the OS - and as noted in the attached screenshot, while only hanging slightly, Parallels was *apparently* using 539.8% of the CPU??? How is that even possible?

 

[AlekseyM]

The % is based on total CPU usage. So 500% means at least 5 cores are being used at 100%. I'd say you are having an IO issue, but you said 1tb SSD (nMP?). During updates and installs, CPU spikes are normal, but quite honestly, open up a ticket for this. If no known fix (I couldn't find any) is available, an agent will connect to your mac and go through logs, etc and escalate to the developers if needed.

 

[PhnordP]

It's a Retina Macbook Pro 1tb SSD. I have not experienced this sort of IO issue with any other app...

I just tried installing a new VM of Windows 7 to see what happens in the Activity Monitor. I ended up having to take a screenshot with my phone, because the computer was so far hung that I couldn't get a screenshot off.

Interestingly, the CPU didn't seem heavily strained...

Upon further inspection, it seems like it's happening when there is heavy Disk I/O AND the Physical RAM is exhausted. I have 16GB of RAM, but when I cut that in half for Parallels to use it leaves my host with only access to 8GB, and it looks like it's using up that 8GB between a handful of applications running in non-intensive ways in the background, combined with the Host OS's disk management, perhaps?

 

[AlekseyM]

If it was a major bug in Parallels, I'd think that everyone would have issues. I have a rMBP as well (similar specs). One thing I noticed is if you give a VM too much memory (or cumulatively if you run more than one) you should never go more than 1/2 of your total memory. What happens when the VM (windows specifically) runs out of memory, it does what's called memory page swapping. Just imagine instead of using memory it uses the hard drive as memory. Even with the worlds fastest SSD it still does not compare to RAM and RAM is still a good 10-20x faster. On top of that, the insane amount of I/O takes up even more memory and Parallels manages that (it's not even a direct disk write).

You'd be surprised how much memory a mac actually uses. I just tried for the hell of it to assign 28 out of 32gbs of my mac pro to my VM and it was crazy. Everything locked up. I actually just forced rebooted the machine. I guess the mac uses all the memory it can get. For instance, right now my Mac pro in activity monitor shows that it is using 15 gbs right now JUST for the mac (out of 32-look at wired memory). Nothing is really running outside of mail, safari, itunes, messaging, etc.

Also don't overdue the cores. Never assign a VM more than the physical cores you have (even though you have 8 virtual, you only really have 4 physical). Also, when it comes to cores, add up all the VM's. You can't do 4 on one and 4 on another, that will unfortunately lock everything up.

Just a few things to try before you submit a ticket with Parallels. I did have REALLY bad experience when my iMac only had 8gb of ram. I even assigned 4gb and it would still run extremely slow. Not sure to blame Parallels on this one or Mac. As soon as I upgraded it to 16 things ran smooth. For whatever reason even assigning 8 to the VM still kept things running smooth. Just my experience, but play around with it and see if you get the same results as I did.