[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Xen and VMware

> Tom Hibbert wrote:
> > Typical corporate talking head blah. Interesting that he talks about
> > Vmotion as a 'killer technology' when Xen is already doing it. 
> > If you read the interview there's quite a bit of Xen-bashing. The guy
> > rags on the requirement for modification of the host kernel. He is
> > comparing the two products as if they took the same approach, failing to
> > mention that Xen's paravirtualisation architecture eliminates all of the
> > costly performance overhead his company's product is famous for...
> I still think you need to respect VMWare for breathing new life into 
> virtual machine research. And I still think Xen can be improved (as is 
> happening now) with regards to memory footprint etc. VMWare ESX is also 
> likely to have some performance benefits from having the drivers in the 
> kernel, even though that comes at the cost of them having to implement 
> the drivers themselves. And the fact that they can host Windows is a big 
> win with lots of customers.

VMWare deserve respect for doing a /heroic/ engineering job; there are
(as we all know) a bunch of really tricky hurdles to get over in order
to do full (or almost full) virtualization on regular x86 hardware. I 
imagine there is some incredibly intricate and hairy code in there to
deal with the binary scanning + rewriting, etc. 

That said, they may be solving the /wrong/ problem. As Xen (and other
systems, e.g. exokernel) have demonstrated, most punters don't give a 
damn if their OS is bit for bit identical or not providing all of their
applications run correctly. Doing all the heavy lifting of full 
virtualization means that you have lots of complex (and fragile) code, 
and also pay a hefty performance cost. It only makes sense if you 
want to play the licensing game... which is kinda up in the air at 
the moment (e.g. if you experience difficulties w/ windows under vmware
you need to reliably reproduce it under a vanilla install too or MS 
won't support it). 

> VMWare does some amount of paravirtualization, with all the VMWare tools 
> that you need to install. In fact, Xen seems to be going in VMWare's 
> direction (shadow page tables, writable page tables, binary rewriting, 
> Vanderpool support) in some areas.


  - our shadow page tables are quite different to VMware's (as far as 
    we can tell - there's no docs on their implementation details) 

  - I don't think that writable page tables are something VMWare do 
    since it only makes sense for paravirtualized memory systems.

  - our binary rewriting is a tiny fraction of what they do and, in 
    current default installs of xen 2.x, is not used at all. 

  - we released VT support back in 2004; I don't believe VMWare 
    ship anything supporting VT at present. 

Anyway, the game may well change with the advent of VT and Pacifica; 
there are certainly both technical and legal/economic challenges to 
be overcome. 



SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.