We have a new server a HP DL360, 64GB RAM, 2x8 core CPU, SAS HDD Raid1 array and one SSD Raid1 array. The SSD Array is configured with 2 Samsung 850 Pro devices (not HP certified).
We installed a Centos 7 OS as a Host OS and we have only one VM with Windows Server 2012 and SQL Server 2012, with only one storage device (qcow2 file on the SSD Array) for the C disk. The Storage and network are accessed via virt drivers. IO performance is very good on the VM too. Memory allocated to the VM: 20 GB. Allocated 10 cpu’s to the VM. We've installed a web app (via IIS) which accesses some 20-30 different companies (same database structure)
The VM runs very well after reboot but in time (50 minutes - 3 hours) we have a big performance degradation. If we restart the SQL Server Service alone, we have another brief dozens of minutes of good performance.
The old old production environment runned well (but only one cpu, only 8 GB of RAM) but yet we wanted a bigger machine! So it’s not clear what happens. When the sluggish behaviour begins on the VM there is no cpu big consumption, just the webapp is not running well. We don’t have a surge in demand. Only one thing seems strange, the memory consumed by the VM never reaches the 20GB limit. The top limit is about 8GB.
We optimized SQL Server, allocated 15 GB (max-memory) only to SQL... runned all kind of testing and diagnosing tools (Toad spotlight for SQL). Everything is on green.
We have no clue to why the server slows down. I assume some kind of incompatibility hardware-kvm-guest VM. Could there be a problem with a sql Server settings when runned in KVM? Or maybe the SSD’s have compatibility issue on the Host? Any ideas are welcomed!
