Win7 Guest OS performance degrades over time

[bazagee]

Hi All,

Looking for some suggestions on how to troubleshoot this (probably well documented some place) issue.
I use this Win7 VM as my primary working OS for my job. The intent was to have a Host machine that I could run as a platform to test configure and deploy VM's from to production.

When I first tired out VBox it was on Mint 16 and using version 4.3 (ish) with a win7 guest. The VM was very fast and snappy. Since then I've moved to Mint 17 with the same vmdk file guest. Everything runs well for about a day then I start to see performance degradation to the point I have to restart the Guest.

The degradation starts with keyboard buffer duplication and triplication when using the info from the clip board. Very slow keyboard buffer response then typing in any application. It then gets to the point of seeing the system hang or freeze while the guest waits for reads and rights. Red dots appear on the HD icon.

I've searched and tried to troubleshoot as best as I can. I've moved the VM guest files to its own dedicated to HD. I've tried removing shared folders and mapped drives. I've looked at Nic drivers and tried changing Network type connections to the host. There and some other reconfigurations to the VM have lead to no resolution.

Thanks for any input.

Current Setup:
Virtualbox 5.1.8 r111374 with the same Extension Pack version On Mint 17
Hardware: i7 processor with 32GB RAM UEFI bios on Asus Maximus ROG motherboard.

[JSA]

Win7 Guest OS performance degrades over time

So not unlike a REAL Win 7 machine then....

Seriously, if the only thing that changed was the underlying Host OS, I'd start searching there.
If you can't see any massive memory utilization in the guest, or any pegged CPU usage its more
likely in the Host.

Do the utilities in the host indicate undue CPU usage or climbing memory usage?

You mentioned putting the VM on its own hard drive, so I assume you might have suspected either contention for the drive with the host or degrading disk performance of the original drive. But if the host is still running from that original drive, it can affect everything.
So many run smartctl on the host, looking for remapped sectors, read errors, etc.

[bazagee]

Thanks for the reply. Sorry I missed the reply email to the post in a timely manner...

I haven't thoroughly investigated the host side but I shell look up the drive performance as you suggest. The host is running off of an SSD that seems to perform well. I never see over use of either CPU or the 32GB of RAM available. I have notices only maybe one or twice where the Mint Host has 'locked up' but that's extremely rare. The host usually just ticks along while VitrualBox do do's the bed...

[WFV]

Have you tried clearing out all the guest cache's and temp's and defragging the W7 guest? As JSA mentioned, it is very like Windows normal performance over time...

[mpack]

It actually isn't true. Windows performance doesn't degrade over time. What does happen is that users install more and more stuff that cruds up the works, especially with apps that install background update checkers etc.

Clearing cache's and defragging are worthwhile but will have a marginal effect IMO. Much more effective would be to view the task list when the system seems slow, and see which tasks are using up all the CPU or network bandwidth. Uninstall as appropriate.

If you search online to find out what a process is, you should of course ignore the sites that tell you you need a virus check or registry tuneup.

[WFV]

What does happen is that users install more and more stuff that cruds up the works, especially with apps that install background update checkers etc.

True, the OS doesn't actually degrade, none of the installed software literally degrades, only the hardware can degrade, electrons have no impact on the lifespan of applications or operating systems. My own experience is I usually get better than marginal effect cleaning the junk out of Windows (cache, temps, and a good reputable defrag utility not Windows built-in) because Windows does a poor job of handling it over time, but as mpack points out, checking the tasks and approaching from there is essential to maintaining a healthy Windows system. Google's background update checker can be a resource hog for example, and Google continues to run in the background when the web-browser is closed, even if its configured to 'not run in background' in its 'advanced settings'.
I've run two of the same Windows guests for the last 6yrs on a Linux host with no performance problems since installing them. More often than not when I have run into guest performance issues it is something wonky in the host caused by recent updates (occasionally VBox updates that are usually easily corrected). I ran Con Kolivas linux-ck custom kernel for a few years until it switched to MuQSS which made XP guest blue-screen and all others including the linux guests crawl such that they were not usable. Switched back to stock kernel and all working fine. CK's MuQSS doesn't support containers (out of the box, or as configurable option per se) - requires further customization to create stub for cgroups, which is over my head presently.

[bazagee]

Thanks for the feed back and discussion. I agree, Windows takes 'time' to degrade as the registry and multiple caches become 'dirty'. What I am experiencing is a matter of days for the degradation. In fact today even after rebooting the Guest I'm seeing an immediate performance issue or buffer response. This is leading me to think that it is something that VB is doing (or not) with the underlying hardware translation. I'm answering this from the Mint 17 host with no sign of performance issue. In any of the resource numbers.

I have a couple of VB guests on the go at the moment building out an Admin Win10 box, and Win 2012 Server image. but only two VM's are actually started at the moment.
So I'm failing so far to see a connection.

Cheers.

[bazagee]

It would seem that my problem was (still testing) related to the on-board NIC and its drivers. Although I had tried several drivers for the Broadcom network chip set the host was never effected.

The apparent fix was installing a Intel Server Class based NIC. I did not install any additional divers at this time wanting to "see" what was going on.