Database Reference
In-Depth Information
Table 10.1
SQL Server Baseline Infrastructure Metrics
While you are collecting and analyzing the various infrastructure metrics, you need to
pay attention to averages and peaks. Ideally you will be capturing the existing system's
performance during a cyclical system peak workload period, which might be end-of-
month, end-of-quarter, or end-of-year processing. You need to determine when the
likely peaks are based on your business and your particular systems. However, if you
are not able to capture the peaks and the averages during a peak business and system
period, you will need to make adjustments based on your knowledge of those times. In
some organizations, peaks of system volumes can range from 20X to 100X normal
system volumes.
If your system is supporting an Internet-facing application, you will have to take
unexpected peak loads and spikes into account. This can be very hard to predict, so you
should take a conservative approach based on what you think the worst-case scenario
is, or the maximum that has been observed in the past. You can then, based on your
workload model and business knowledge, extrapolate what may be required, taking into
account known future activities.
Note
A vCPU of a virtual machine is only single threaded. Unlike a physical
CPU core, it does not include a hyper-thread. If you want to get the most out
of your system and do like for like comparisons with the physical source
