On April 30th, along with Chris Carter from Cisco, we delivered a technical webinar covering best practices for implementing XenDesktop on Cisco UCS Platform. This was the first of our TechTalk series with Cisco and in the coming months we will deliver further on our joint solutions. So if you have interest in a specific topic or are working on our joint products deployed in production, please leave a comment with future topic suggestions.
To the 235 people who attended the webinar, we thank you for your time and hope it was a value for you. As mentioned during the call, the summarized Q&A is listed below.
Here is the recorded webinar link for those who didn’t get a chance and for all geos since this was covered during Americas time zone. For the next one, we will have it for all 3 major time zones.
Here is the link to the recorded video we played on UCS PowerShell script for configuring service profiles with these XenDesktop best practices in mind.
And off course the webinar was based on the following paper Chris and I wrote a few months ago:
Summarized Q&A from the Webinar:
Q. Does SR-IOV supported for use as HBA with Citrix XenServer?. There is no cisco CNAs included at the hardware compatibility list for XenServer documents only talk about SR-IOV for NICs, not for HBAs?
A. Cisco currently only supports Hyper-V and KVM with SR-IOV.
Q. What teams need to be involved in making this changes? e.g. Network, SAN etc.?
A. As with any deployment of a server, you will need input from the Network, SAN, Security, and other teams. The difference with Cisco UCS is that you will only need detailed input one time, while creating the Service Profile Template, and after that you’ll simply be deploying from the template.
Q. Do you recommend leaving HyperThreading off?
A. No, we recommend that you enable HyperThreading (by default it is enabled).
Q. Are these for XenDesktop/XenApp running on XenServer only or does this apply to ESXi as well?
A. We built the document and the script to support XenServer, ESXi, or Hyper-V. Some specific considerations need to be taken for respective hypervisors. For example, while designing the VNIC template, and creating VLANs for XenServer based deployments, your MGMT and XenMotion networks would reside on the same NIC as they cannot be split. This is not the case with vSphere where you could separate them on different interfaces.
Q. How many virtual interfaces can a VIC support? Is this similar to HP’s Flex 10?
A. Much higher number of interfaces than Flex 10 – 256 virtual adapters with the current VIC (theoretical – supporting over 100 today) – http://www.cisco.com/en/US/prod/collateral/ps10277/ps11551/data_sheet_c78[/url]-677682.html Flex 10 supports 4 interfaces. The VIC also supports HW Fabric Failover independent of the OS. HP would rely on OS based NIC teaming.
Q. If I split the VICs as 4 logical NICs, is the bandwidth assigned to each static or it is dynamic based on workloads
A. It is dynamic, used as needed. If you want to guarantee that any particular NIC gets a certain percentage of bandwidth, you’d utilize a QoS policy and assign it to the NIC.
Q. Is Cisco office in a box part of the cisco UCS?
A. Cisco “Office In A Box” leverages the 2900 and 3900 Cisco Integrated Services Router and the UCS E-Series blades to host Citrix XD, XA, and other infrastructure services in virtual machines.
Q. Is the Cisco “Office in A Box” solution able to scale for large # of users say 2000?
A. The office in a box solution is ideally suited for very small deployments – around 25 users (up to 40 I believe). The UCS E-Series blades used in the Cisco ISR do not have enough memory to host more desktops than that. If you are instead referring to “VDI-In-A-Box”, leave a comment.
Q. Does the Cisco UCS come with Fusion IO and/or are there any diskless VDI implementations that can be done?
A. Fusion-IO is optionally available for use with Citrix XenDesktop. It is ideally suited for non-persistent desktop environments (pooled) where changes made to the desktops do not need to be maintained between logins/logouts or failures.
Q. Any advantages of boot from san vs. from local disk?
A. Yes, utilizing boot from SAN (or boot from iSCSI or PXE boot) allows the server to be completely stateless, allowing the movement of the service profile without concern for the movement of the data on the local disk.
Q. When will Citrix From Wow to How occur this year?
A. There will be an update on that soon, stay tuned to Citrix for more information.
Q. We found that the turbo boost option was not in the bios of the B200M2
A. Turbo Boost was introduced by Intel starting with the Xeon 5500/5600 series processors, which the B200 M2 utilizes. It is enabled by default.
Solutions Architect, Citrix Systems