Database Reference
In-Depth Information
production. When using this setting, you should configure reservations at the virtual
machine level (previous section) to prevent the vSphere balloon driver from
encroaching on the SQL Server's memory footprint. The reason for this suggestion is if
memory is locked by SQL Server and the vSphere balloon driver kicks in, there can be
a negative impact to the performance of your SQL Server virtual machine.
For Tier 2, Tier 3, quality assurance, test, and development systems, we suggest a
review of this setting prior to implementation. If consolidation and memory optimization
are the desired goals for one or more of the aforementioned tiers, then do not implement
the Lock Pages in Memory setting, because this will have a negative impact on the
performance of the SQL Server if the vSphere balloon driver (vmmemctl) kicks in and
attempts to reclaim memory from these virtual machines. As discussed in the “ Large
Pages ” section in this chapter, when large pages are configured for SQL Server, the
default behavior at boot time is to find contiguous pages in memory and claim these,
thus affecting the overall memory footprint for SQL Server.
Storage
How you configure and present storage to the SQL Server virtual machines will have a
profound impact on their performance. Chapter 6 , Architecting for Performance:
Storage ,” goes into immense detail around storage configuration, so we will only
discuss the highlights in this chapter.
First, remember that the rules used to size SQL Server in a physical world carry over to
the virtual world—they just need to be tweaked. Too often we see customers just pick
up SQL and throw it over the proverbial wall. Why is this? Well, let's take a trip down
Memory Lane; let's look back at the last 10 years and compare virtualization and
database implementations.
When we look at virtualization and how virtualization became mainstream, we notice
that the first systems virtualized were test and development systems. These systems did
not impact the business, and if they went down or had poor performance, only IT
noticed. As the software matured and the hardware evolved, we saw departmental
applications go onto the platform, starting with those owned by IT, and eventually
moving outward to non-IT departmental applications. And here we are today, where
there are few workloads that exist that cannot be virtualized due to the work by
VMware, independent software vendors, and hardware vendors. However, what we
didn't see keep up was the systems running vSphere, particularly the storage
subsystems.
Now, hold that thought for a second while we examine the database trajectory. Some
argue that data is the lifeblood of any company and that the health, performance, and
availability of databases are a reflection of how much a company relies on this data.
For a large number of companies, this was important, so when the DBA team said they
 
Search WWH ::




Custom Search