Databases Reference
In-Depth Information
if need be. However, the test servers, while only occasionally used, might have a higher priority
than the development servers, and therefore have a weighting lower than the production servers but
higher than the development servers.
IDENTIFYING CANDIDATES FOR VIRTUALIZATION
As virtualization's popularity has grown, so has the debate about which server roles can be virtualized.
Some of the very broad generalizations have contained a grain of truth. Certainly for a long time
there was the view that production database servers could never be virtualized but virtualization
technology has developed signii cantly to not normally make that a problem. In fact, VMware now
suggests that virtualizing 100% of an enterprise server estate is now a possibility with the latest
versions of their software.
Ultimately, the only way to determine how much of your server estate can be virtualized is to adopt
some high-level guiding principles before performing a detailed review of relevant performance data
to ensure your virtual environment can handle your anticipated workload.
Guiding Principles
When people ask me which servers they can and can't virtualize, their question is often “what's
the biggest server you'd virtualize?” My answer is always that far more factors should inl uence the
decision than the server's size alone. This section outlines some general guiding principles that you
can follow when considering virtualization. More detailed information is provided in the next section.
Server Workload
The i rst area is related to the server's workload. Although you might assume that this is the same as
the server's size, a small server working consistently at its highest capacity may be harder to virtual-
ize than a larger server that is often relatively idle.
Today, with a well-designed virtualization environment it's safe to work to the following design
assumptions:
➤
Quiet
server workloads can be and are routinely virtualized today. These might well be
domain controllers; i le servers; or the database servers for your anti-virus software, your
expense tracking, or your HR system in a small to medium-size business. If the server is one
whose performance doesn't require monitoring or there is no question about it always meet-
ing its business requirements, then you can consider this a small workload.
➤
Active
server workloads also can be and are often routinely virtualized; and as long as
capacity checks for the required CPU, memory, and storage throughput are made i rst, these
workloads usually virtualize well. This kind of server might well host your i nance system's
database server, where for the majority of the time a constant number of data entry users
execute low-cost queries, some utilization spikes occur with ad-hoc reports, while month-end
processing creates an infrequent but demanding workload. You may well already be using
some performance-monitoring tools to proactively look for slowdowns in the end user
experience, as well as perhaps deploying a physical server to which you know you can add
more CPU or memory as needed.
Search WWH ::
Custom Search