Our team in the Solutions Lab recently conducted a series of single server scalability (SSS) tests comparing performance among storage devices. In this case we focus on local solid state drives for write cache storage.

Citrix XenDesktop hosted shared desktop allows the creation of sites to support users with multiple hosted shared desktop servers providing resiliency.  These servers can be virtualized and streamed through Citrix Provisioning Services to manage the server image.  Provisioning Services requires a file cache, referred to as the write cache, and is often stored on SAN storage .  The file content gets deleted and re-created on every hosted shared desktop virtual machine during boot and contains temporary data.  We decided to examine this scenario using  local solid state drives on a single Cisco UCS blade to store the write cache files and take advantage of the Citrix XenDesktop site design for resiliency.  In addition, we validated that the local solid state drives satisfy the write cache I/O requirements.

Let’s talk about our test environment components, configuration and test results.

These are the hardware and software components used for hosting the hosted shared desktops: Cisco UCS B200 M3 blades with Intel® Xeon® E5-2697 v2 (“Ivy Bridge”) processors, 384 GB (1,333 MHz) memory and two 372GB SSD drives (raid 0 striped UCS Configuration Policy), XenServer 6.1.

The virtualized delivery infrastructure was hosted in a separate XenServer pool as well storage that includes Citrix XenDesktop 7.0, Provisioning Services 7.0, Citrix StoreFront 2.1, Login VSI 3.7 software from Login VSI Inc. (http://www.loginvsi.com), user profiles and vDisks storage.

Hosted shared desktop configuration

The hosted shared desktops as a virtualized streamed VM (through Provisioning Services) solution has the advantage of leveraging local solid state drive tests. The space requirements for hosting write cache files is minimal, while the user density is higher when compared to hosted virtual desktop.

The hosted shared desktop streamed master image has the following configuration: five vCPUs, 24 GB RAM, 80 GB vDisk size, 25 GB write cache, one virtual NIC, Windows Server 2012, Remote Desktop Session Host (Windows Server 2012 server role), Login VSI 3.7 target software, Microsoft Office 2010, Citrix XenDesktop VDA agent and Citrix Provisioning Services target device.

We used Provisioning Services 7.0 XenDesktop Wizard to deploy eight hosted shared desktop virtual machines.

Test results

Hosted shared desktop test results show similar density between storage types. This confirms that we can run the load typically hosted on enterprise storage on local solid state drives on the Cisco blades, provided there is suitable I/O performance for the write cache.

Each test was executed and validated multiple times to ensure consistency.

Hosted Shared Desktop on Remote Storage – VSI Max Result (VSI Max corrected=255)

Hosted Shared Desktop on SSD Storage – VSI Max Result (VSI Max corrected=253)

Some advantages of using local solid state drive storage over remote storage (SAN) for write cache are costs savings while providing the required I/O performance. In addition, we found that the session density is very similar between local solid state drives and remote storage.

To improve redundancy when using local solid state drives storage, we suggest you use raid mirroring with hot-spare, and XenServer Storage XenMotion.