[Hyper-V] Too Much Hardware Resources?

Question

Hello,
I have a Small Business Server 2008 standard (SBS2008) install that is
virtualized. The virtual was installed and set up on Hyper-V Server 2008 originally 5 years ago. I simply copied the virtual server from 2008 to a new set of hardware running
Hyper-V Server 2012 R2 with NUMA virtualization enabled. 
The new hardware is a Dell PowerEdge R520 server with dual Xeon E5-2420 2.2 GHZ processors, 96 GB Ram (16384MB @ 1600 mhz x 6 DIMMs spanned across 2 NUMA nodes), and 6 600GB Seagate Cheetah 15K RPM SAS 6Gbs drives in a RAID 10 configuration on a Dell RAID
Perc H710P (which is made by LSI)  and we know that our disk I/O benchmark tops out at
3.5 Gbps read and write on the appropriate file sizes that they would typically read and/or write. The NICs are dual gigabit Broadcom NetXtreme NICs teamed together through powershell (http://technet.microsoft.com/en-us/library/jj130847.aspx)
In short, the new hardware is FAST. Really fast.
On the SBS2008 install, we initially assigned 18 logical processors and 24 GB of ram because we thought let's throw as much resources to it as possible to ensure it is also fast. 
Up until yesterday, nobody was having a problem, for approximately two weeks. Then for no reason, the server begins to perform horribly - pings to itself, from itself, would avg around 80 ms, spike as high as 300 - 400 ms and not go below 10 ms. This should
always be <1 ms. Pings to the Hyper-V server from a workstation were consistently <1 ms. Pings to the other virtualized server (Server 2008 R2 Enterprise running remote desktop services with 56 GB ram assigned and a lot of processors) were also consistently
< 1 ms. From Hyper-V to the SBS2008 virtual were also as bad as pinging SBS2008 from itself to itself. Not until I was able to stop Microsoft Exchange and SQL Server did the server start to respond 'normally' again. I then shut down the server and decreased
ram to 12 GB and logical processors from 18 to 8. I also disabled NUMA virtualization. Since yesterday at 3 pm, the server has been performing great. 
Questions:
1. Could it be possible that we had assigned too many logical processors or RAM initially to the SBS2008 virtual guest?
2. Could it be possible NUMA virtualization was causing these issues?
3. Is it possible the issue could still not be resolved and would represent itself again in two weeks? 
4. This SBS2008 virtual guest is a 5-year old install, installed according to Microsoft's best practices guide on SBS2008 installation. Beyond that, could there still be some configuration issue that could result in that kind of poor performance?
Any other thoughts and suggestions are welcome. The server is again performing fine since the changes mentioned above, but there was no definite resolution discovered, so I have no proof that it is resolved until time tells us otherwise.

Answer 1

#1 is your likely cause. What you're really assigning a slice of CPU time, not a true processor.
Excessively assigning CPU resources to a VM rarely results in better performance and can definitely result in worse performance for that VM and others as well. The maximum
supported processors for 2008 is 4.  Going beyond the supported and tested processors can result in unwanted behavior.
For other OSes, look here:
http://technet.microsoft.com/library/dn792027.aspx
Also, for future Hyper-V questions, a better place to post is the Hyper-V forum, here:
https://social.technet.microsoft.com/Forums/windowsserver/en-US/home?forum=winserverhyperv
It's where all the Hyper-V gurus hang out.
This one is for the product Virtual Server 2005.