An exciting day today: The Wind River Hypervisor has been officially released. You'll find a lot more information on the announcement web page.
This is a very significant release, it provides embedded system developers new capabilities to partition their systems, be it single core or multicore. Customers have been requesting these capabilities as they are planning their new product lines, for the short, medium and long term roadmaps. Multicore is hitting the main street and Wind River is ready for it.
Have a read through the material provided, if you are itching for a demonstration, then have a look at these videos:
Mark is a senior product manager with Wind River focusing on multicore and virtualization solutions. Prior to joining Wind River Mark has helped development teams build embedded systems across Asia, Europe and North America in automotive, telecom, consumer electronics and defense industries.
Subscribe to RSS feed.
Hi!
1. Where did you get this technology? Is it Xtratum? Is it related to Xen?
2. What is the performance penalty of running a) linux and b) vxworks under the hypervisor or running them by themselft on the hardware?
3. Does the hypervisor have a name?
Posted by: Jan Tångring | June 16, 2009 at 07:36 AM
Hi Jan,
Thanks for your questions. The 'Wind River Hypervisor' (that is it's name) has been developed from scratch internally at Wind River. It is not related to Xen and it is quite a lot thinner and more responsive to Xen (based on customer feedback). The Wind River Hypervisor is custom developed for embedded systems.
The performance penalty depends on the processor architecture and the workload, but typically under under 2-3% and often under 1%. This is comparing a native Linux/VxWorks to a hypervised Linux/VxWorks.
The Hypervisor is available for evaluation if you are interested. You can contact Wind River in your area and they will happily assist you.
Regards,
Mark
Posted by: Mark Hermeling | June 16, 2009 at 08:16 AM
Thanks
Posted by: Jan Tångring | June 16, 2009 at 10:10 AM
Are those performance numbers for an x86 with vt support running a single guest per core?
There are a multitude of combinations of PowerPC, x86 (with or without vt support), running a single guest os per core, running several guests on a single core - would all these be having the 1-3 percent performance?
Posted by: Jan Tångring | June 16, 2009 at 09:25 PM
Congratulations Mark. :) I've posted some thoughts to my blog...
Posted by: Hollis Blanchard | June 17, 2009 at 09:04 AM
Hi,
I was reading your other post:
http://blogs.windriver.com/hermeling/2009/04/cavium-octeonii.html
Hypervisor support for Octeon. We are interested to know more about. Can you please connect me with your representitives?
Thanks,
Posted by: Saikrishna | June 23, 2009 at 09:18 AM