Database Reference
In-Depth Information
Chapter 9. Architecting for Availability: Choosing the
Right Solution
There are many choices when it comes to protecting your database, but the database is
not the only application to consider. Too often individuals become too focused on a
technology and not a solution. This chapter will not provide a “silver bullet solution”
for database availability. The goal of this chapter, and the philosophy of the authors, is
to provide the right solution for the right use case based on customer discussions.
Although there are many “right” solutions, it is important to design the right solution for
the right use case.
This chapter will walk through options available to DBAs and vSphere administrators
for high availability, business continuity, disaster recovery, and backing up their
databases and their infrastructure. What value have you provided the business if the
database is up and running but the rest of the application stack is down?
Determining Availability Requirements
It should not come as a surprise that one area of database virtualization that can lead to
hours of discussion, with both vendors and internally with the customers, is protection
of the database. DBAs have been entrusted with protecting the company's most
important information. It is often said that the database is the heart of a company. Lose
the wrong database, and the company cannot make business decisions. So, naturally,
DBAs tend to be very conservative and will go with a solution that provides the highest
availability and provides them with granular data protection. And why not? Lose the
data, lose your job.
However, what about the rest of the application stack? Does it matter if the web servers
are down but the database is available to accept incoming requests? Actually, years of
experience and meeting many customers has proven the answer to be a resounding, “It
depends.” It depends because for some customers, DBAs are measured and
compensated independently of the web and application tier administrators. For these
DBAs, we have found their primary concern is making sure the database is available.
And this is okay, because by placing a database on ESXi, we can make it more
available than in the physical world.
For other customers—and we have found this to be an increasing trend—the stack is
being measured for availability. Therefore, the entire supporting cast is measured
against whether the application is working or not working. You either get full credit or
no credit. This ensures that all teams are working together to support the business. The
business only cares that the entire stack is working, not your individual part of the stack.
Sounds a bit harsh, but this is the truth.
