Dell EqualLogic FS7500 - Manual (Page 11)

Dell EqualLogic FS7500 – Unified block and file storage for virtual workloads
Page 9
Together these metrics can be used to quantify performance at the application level. Both of these
performance metrics are related to each other. When the storage system reaches its performance
limits, the transactional performance and latency values are often found to be in an inverse
relationship. As latency increases, transactional performance decreases because of the increased time
required to process each I/O. However, when the system is underutilized, transactional performance
can increase without dramatically affecting latency. As a result, when assessing the performance of a
storage system near saturation, it is sufficient to measure the application transactional performance at
an acceptable quality of service (QoS) level that is typically defined by a maximum application
transaction response time.
Test Methodology
Our test environment was structured to provide insight into the scalability of VMware virtual
environments supporting multiple simultaneous workloads and VMware-specific virtual infrastructure
operations.
In this test environment, we deployed a pair of FS7500 controllers and the VM host servers with
sufficient CPU, memory and network resources to avoid any bottlenecks in these subsystems. In terms
of hardware, we varied only the storage configuration to understand the impact of increasing storage
capacity in the configuration. For each storage configuration, we increased the workload and noted the
application transaction throughputs and response times. In our test results, we focused on quantifying
application transactional performance while maintaining a steady virtualization infrastructure activity
as well as maintaining an acceptable application QoS.
Figure 3 depicts a VMware environment with multiple virtual workloads hosted on an EqualLogic
FS7500. This is the overall layout of the test environment.