Databases Reference
In-Depth Information
MONITORING VIRTUALIZED DATABASE SERVERS
Hopefully the previous section has given you sufi cient guidance to architecting and deploying your
i rst virtualized database servers, even if only in non-production environments. This section focuses
on real-world monitoring of your virtualized database servers, identifying specii c changes you may
need to make to your monitoring processes and coni rming which aspects of your current
monitoring can remain unchanged.
Traditionally we have monitored Windows servers and servers running SQL Server with tools such
as Performance Monitor. These tools have counters that are designed to expose the true utilization
of a server's hardware and the operating system's demands on it. For example, we can look at the
workload of a server's CPUs by monitoring the % utilization values shown in Performance Monitor.
Likewise, we can see how much memory the server has both used and available by looking at
similar counters. These counters were perfect in the physical server world because we knew if
Windows booted up and saw 4 logical CPUs and 16GB of memory then all of that resource would
be available to the operating system and usually SQL Server as well.
This can cause issues; what does 100% of CPU utilization or 8GB of available memory actually
represent in the virtual world? In environments where no restriction, contention, or over-allocation
of resources has been coni gured, some certainty can be found from performance data. In larger,
more complex environments, contention ratios or memory allocation might be changing on a
minute-by-minute basis.
The example shown in Figure 17-7 demonstrates how in a Hyper-V environment, the same
performance metric monitored in two places can be so different because of an underlying resource
limitation in place. VMware provides its own Performance Monitor counters through the
VM Memory and VM Processor objects.
Processor
Processor
Host
Guest
Guest
% Processor Time
% Processor Time
OS
OS
OS
Hyper-V Hypervisor
Host
Logical Processor
% Total Runtime
Hypervisor
Per Guest
Host Server
Virtual Processor
% Total Runtime
FIGURE 17-7
Search WWH ::
Custom Search