It’s 2015, happy new year! This is going to be a big year and not just for the XenServer product. As everyone gets back into the swing of daily commutes following the Christmas and New Year holidays, I along with the rest of the Citrix XenServer team are busy getting the latest release of XenServer ready for shipping.
I have recently taken over the Product Marketing Management role for the XenServer product line and there’s plenty of work to be done as we ready the product for release, including blogs, FAQs and data sheets to name a few. It’s also an exciting time, as this release will introduce some great new product features and enhancements, and the Engineering and PM teams are already busy working on delivering the great roadmap for future product releases later this year.
Regarding XenServer v6.5 I want to highlight to you why this is a significant release for the product and why you owe it to your business to check it out?
The past few releases of XenServer have seen some large scalability improvements; however, this was starting to stretch the capabilities of the underlying architecture. New and more scalable hardware means customers trying to get the best return on their investment. XenServer needs to be able to scale, and not only in numbers of virtual machines, but also in being able to handle that volume of VMs’ IO.
In XenServer v6.5, some of this architecture has been updated; there are some new and reintroduced features, product performance enhancements and some new pricing and packaging.
64-Bit Control Domain
The control domain, which now utilizes CentOS 5.10 is 64-bit. This removes the previously restrictive low/high memory division of a 32-bit domain and allows for more PCI devices and memory. It is automatically configured in dom0 based on the amount of available free memory on the system.
Latest Xen Project Hypervisor v4.4
Amongst the Xen 4.4 hypervisor improvements is an increase to the number of event channels, particularly important to some desktop virtualization use cases where there are high numbers of in-VM devices (disks, network adapters, etc.). Additionally, the use of grant-copy has dramatically increased aggregate VM disk and network IO performance.
Following on from the Intellicache feature introduced a few years ago, the in-memory read-cache enables the ability to store gold-build image data locally to the host in RAM, dramatically improving boot and run time performance of XenApp/XenDesktop use cases.
Previously, in order to handle high volume IO scenarios, in part generated by large numbers of VMs per host, a larger, more expensive array would have been required to handle the IO volume. The in-memory read-cache means more modest arrays can handle the same, if not more, virtualized desktop IO than more expensive options without it. This is great for both greenfield desktop virtualization projects and scenarios where existing infrastructures are being expanded for new additional load and/or use cases.
- Reducing off-host IOPS via a host-side in-memory read-cache
There are several improvements, including the ability for XenServer to free up previously used blocks from a LUN (i.e. deleted VMs from an SR), live LUN expansion, support for tapdisk3 and an updated device mapper multipath.
There’s an updated Open vSwitch (now at v2.1.3), GRO is enabled by default and now a Netback Thread per VIF, which is an improved way of sharing resources and improving performance with para-virtualized VM NICs. There is also an updated DVSC, the distributed virtual switch controller with a number of platform related security fixes.
Citrix XenServer has led the market in its support for virtualized graphics processors. GPU passthrough has been around for a while and even some of our competitors also feature it, however it was never a scalable solution for large volumes of desktop virtualization users. vGPU, introduced a year ago by Citrix with Support for NVIDIA GRID vGPU cards meant suddenly, higher numbers of users could get access to advanced graphics in a virtualized desktop scenario.
In this release, we’ve scaled our support for vGPU sessions by 50%, now with 96 GRID vGPU sessions per host. Obviously you need to have the appropriate h/w host to support 3 NVIDIA GRID cards, with all necessary resources to run the numbers of VM, however, it demonstrates how Citrix continues to lead the market for organizations seeking application and desktop virtualization solutions.
- 50% increase in GRID vGPU sessions with this release
The march of vendor OS improvements continues and as such, there are new guest OSs supported in this XenServer release from vendors including Ubuntu, SLES, CentOS, Redhat and Oracle. New for this release is also the support of some Linux guests running in HVM mode.
There’s the workload balancing feature that has been updated for this release and lastly, but certainly by no means least, there are improvements to our Intel TXT support which now includes asset tagging. There are improvements to Citrix XenCenter with new views and a unified UI and improvements to the SDK.
With the release of XenServer v6.5, Citrix is introducing an Enterprise Edition to complement the existing [Standard] Edition. Both editions continue to be made available on a per-CPU socket basis (of the host) on either an on-premise subscription (a 1 or 3 year term), or as a perpetual license.
Additional features in the XenServer Enterprise Edition include:
- In-memory read-cache
- Dynamic resource scheduling (workload balancing)
- Secure measured boot of the hypervisor platform (Intel TXT)
- vSphere workload conversion manager
I’ve only touched on some of the features and improvements in this release, so please check out the links below for further information:
- XenServer v6.5 release notes found here
- XenServer v6.5 configuration maximums document found here
- Technical XenServer FAQ found here
- Licensing FAQ found here
2015 will certainly be an exciting year for the XenServer product! You owe it to your business to understand how Citrix XenServer can help you to improve business agility and reduce infrastructure costs for a variety of virtualization use cases.