White Papers

BP1018 Sizing and Best Practices for Citrix XenDesktop with Dell EqualLogic Storage 17

5 Citrix XenDesktop test results and analysis

This section describes the different XenDesktop VDI characterization tests conducted and also the key

findings from each test. The Task worker user type represents the majority of VDI users in the industry

today and we focused our testing on this workload profile.

For all single array tests, we used eight M610 blade servers to host 630 virtual desktops. We uniformly

distributed around 80 desktops on each blade M610 server so that server performance limits were not

reached at any point. For two array tests, 16 M610 blade servers were used to host 1270 desktops.

5.1 Test scenarios

We tested the following scenarios:

• Task worker – 100% Pre-Booted

In this test we pre-booted 630 virtual desktops and allowed the desktops to be in idle state for

more than 20 minutes before starting the Login VSI workload to simulate task worker profile.

• Task worker – 20% Pre-Booted

There are many instances in the industry where pre-booting all virtual desktops might not be

possible due to infrastructure resource constraints. So in these scenarios, a minimum number

of desktops are pre-booted and the remaining desktops will boot as the user logs in. In this

test, only 126 virtual desktops (20% of the 630) were pre-booted and the remaining desktops

were booted on demand while starting the desktop sessions.

• Virtual desktop scaling with two arrays

In this test, we wanted to validate the scalability of the VDI solution by adding an additional

storage array. To support 630 virtual desktops, we used eight blade M610 servers and one

PS6010XVS array as explained in the previous section. We added one more PS6010XVS array

to the same pool and also eight more blade M610 servers to the ESXi cluster hosting virtual

desktops and tested 1270 virtual desktops.

• Boot storm

The boot storm represents the worst-case scenario where all the virtual desktops are started at

the same time and they all contend for the system resources simultaneously. We wanted to

evaluate if the storage array hosting the desktops is capable of handling this kind of huge spike

in I/O within acceptable latency limits.

• Login storm

the same time. Login storm scenarios are much more critical compared to boot storms in a

VDI environment. The boot storms can be slightly reduced by pre-booting the virtual

desktops. However, login storms need to be handled by the underlying storage efficiently,

because the login storm is difficult to prevent in a user environment.

All the virtual desktops were pre-booted and then a login storm was simulated by

simultaneously logging into all desktops using Login VSI.