BMD software and virtualization via VMWare vSphere 4.0

In this description, we explain the difference between a physical operating system installation and an installation in a virtual environment. The test system used was configured as follows: 

• Server: HP ProLiant ML350 G5
• Processor: Intel Xeon E5310 1.6Ghz QuadCore
• Memory: 7 GB PC2-5300 DDR2 667 Mhz ECC
• Hard drives: RAID10 from 6 x 72 GB 10k rpm SFF
• Storage Controller: Smart Array E200i with 128 MB incl. BBWC 
• Smart Array P400 with 256 MB incl. BBWC 
• Operating system: Windows 2008 R2

The test scenario consists of three parts. In each test, one run was completed with the weaker controller E200i and a run with the stronger controller P400. The performance was tested with Microsoft's benchmark tool "SQLIO". Only the random accesses (random I/O) were measured. The transfer rate (sequential I/O) is not valid enough for a practical comparison and was therefore not listed. For comparison, a BMDNTCS database update was additionally performed in order to determine the dependency of processor and memory. 

Scenario 1: Physical installation with two partitions (C:\ and D:\)

E200i Controller: 

P400 Controller

Here it's already clearly visible that a better hard disk controller can result in a signifcant increase in speed, while using identical hardware. 

Scenario 2: Virtual installation with one disk and two partitions (C:\ and D:\)

E200i Controller:

P400 Controller

When you compare the physical installation from scenario 1 to the virtual installation from scenario 2, it is apparent that I/O performance of both read and write access is almost equal in the two scenarios. With small block sizes it is even slightly faster. VMWare's proprietary file system VMFS thus seems to scale very well with the cache of the controller. 

Only the duration of the database update is slightly longer than with a physical operating system. This suggests decreased performance of processor and memory due to the hypervisor. 

Scenario 3: Two-Disk Virtual Installation (C:\ and D:\ )

E200i Controller: 

P400 Controller

When using an additional disk, the disk performance in the virtual environment is even better than in both previous test environments. The duration of the database update is also slightly shorter than in Scenario 2. 

Conclusion: 

The statement that installations in a virtual environment are slower than physical installations cannot generally be confirmed to be correct or incorrect. 

Based on the results of these tests, however, it can be said that under VMWare only very little power is lost when using powerful enough hardware. In fact, I / O performance on its own is much better than in a Windows installation. 

Considering the many benefits of virtualization, we can safely recommend the use of VMWare with properly dimensioned server hardware.