We already have "turbo-charged" Your memory throughput with the new CiRwOtD feature (if you are not that familiar with the acronym, you'd better ask somebody!) but what PVS merge, boat and streaming times about improving
this is what I now want to share with you exactly - partly because this configuration can improve performance streaming, but also because this is a "hidden" or undocumented feature of PVS 7.7.
Before I to the point, I just wanted to give a shout-out to the PVS team. PVS 7.7 was released a few weeks ago, together with XenApp / XenDesktop 7.7 and I do not think the hype deserves it - there are some fantastic things we have included in this release. My favorite features are likely to do a real Posh interface and the ability to in-place upgrades, been honest both things we have for a long time begging now. But my other 2 favorite things are more performance-related - VHDX support and multiple threads read IO. Dane Young has done an amazing job to document some of the VHDX kindness and how it can improve things like merge operations (it's a must-read!), So I will in this article focus on the latter.
Turbo-charging PVS Streaming
We snuck something in the 7.7 code that actually improve target device streaming. And it is streaming performance remains poor was not that, it's just better now. Why? The key PVS driver who takes over after the bootstrap is now optimized for multi-core systems. In other words, that the PVS driver does the streaming in the steady state is multi-threaded and can simultaneously process multiple read operations.
The beauty of this new feature is to have literally nothing to do in order to - by default it is "released" is, if you will, but there are two requirements:
- you need the 7.7 PVS to use driver. Everything, which means that you need to update your target devices 7.7. It is important to note that you need to build your BDM ISOs or partitions to take advantage of this you do not re-create (I this point because I had a mistake in the first version of this article, say that you use the 7.7 needed to read why that is)
- you must ensure that your 7.7-based target devices have assigned at least 2 vCPUs or cores - bootstrap .. an indication of this - even if you specify 4 or 16 vCPUs for target devices, we will only assign 2 reading thread max on the PVS page
So if you. with 7.7 target devices and enter 2 or more CPUs, automatically use PVS 2 threads for read operations.
Impact & Results
I have run a few tests already, and the results are impressive. (7.6 with a single CPU PVS), , it will be processed to boot lasted 20 seconds and fully register a target device with 100% of GPOs about 1 minute ago ; it now takes only less than 1 minute (59 seconds on average). And all I did was to give my objectives update to 7.7 and 2 vCPUs. And it is 27% faster
Another thing I need to clarify, because people began to speak about -. You need the "msconfig" tweak apparently do not increase to CPUs for the boot cycle. And you will see more performance improvements in the steady state compared to boot. Why? The reason is that the bootstrap, which will initially be used by the BIOS, can not take advantage of multiple cores -. Only the PVS driver who takes after the bootstrap can
use multiple cores and benefits of this new feature
My former colleague, Dan Allen (now at Bromium, but still a PVS fanboy and we are also on a new White Paper together with team VRC to show the impact of web browsing), and a test was run in its own laboratory and found his load times improved by 25% . Your mileage will vary, of course, but these first numbers appear to be fairly consistent. And this is one of those no-brainer things to implement (no drawbacks or side effects) update simply to 7.7, if you so choose.If you are able to do some comparative tests, as we have done (if it is this new feature or VHDX based), I would like to hear under about your performance gains in the comments section. Hope this trick helps you on the move.
-Nick
Nick Rintalan
Lead Architect, Citrix Consulting Services (CCS)
0 Komentar