Database Reference
In-Depth Information
architecture section. vSphere provides functionality that helps automate the process of
resource allocation to ensure databases get the resources they need, when they need
them, and in the amounts needed. An example would be vSphere DRS. From a
performance perspective, individuals will segment out, if possible, the OLTP, batch,
and DSS workload types. The reason for this is multifaceted.
One reason, which is not often considered, is that OLTP and DSS can (and often do) run
simultaneously during production hours, whereas batch workloads run during off-
production hours. Another reason is OLTP transactions are shorter in duration but
higher in frequency, with a mixed read/write ratio. OLTP architectures focus on
delivering the highest possible IOPS and throughput to avoid disrupting business-
critical operations.
DSS systems are optimized to fewer but longer running queries that are read intensive.
Architects will design these systems to retrieve and present data as fast as possible to
the end user's application.
With OLTP and DSS running concurrently and executing different query types and
different read/write ratios, architects will segment these systems. The last thing an
architect wants is for a long-running DSS query to affect their revenue-generating
customer-facing application.
When comparing OLTP workloads against batch workloads, we notice their activity
levels occur during different part of the business day. Put another way, OLTP runs
during production hours whereas batch workloads run during off-hours. Because of this,
we are able to leverage the power of virtualization and place both of these workload
types onto the same physical host. This is industry specific. For some industries, we
find the previous statement to be accurate, but for others this is not the case. Therefore,
you need to understand the workload types and patterns associated with those workload
types.
Figure 5.3
displays an OLTP-only database server running production workloads during
production hours, considered to be from 7 a.m. to 6 p.m. (18:00). The average workload
utilization throughout the entire day for this server is 21.9%, with 36.8% during
production hours and peak utilization of 53% occurring at 10 a.m. During non-peak
hours, average utilization is 7.1%, which lasts for approximately 12 hours. Translation:
50% of the day the system is working, and 50% of the day the system is not working.
