There's a fundamental structural difference between VMware and Hyper-V. Tim Antonowicz, Senior SE and virtualization guru at Mosaic explains the difference, but is he right?
Update : Read Matt McSpirit's comment.
Saturday, 11 April 2009
Architectural difference between VMware vs Hyper V explained?
Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
Hi Eric,
Where does this chap get his info from? Hyper-V/ XenServer aren't hosted hypervisors! You can read here: http://msdn.microsoft.com/en-us/library/cc768520.aspx for Hyper-V and here: http://www.intel.com/technology/itj/2006/v10i3/3-xen/4-extending-with-intel-vt.htm for Xen/XenServer and see that both hypervisors are running as the lowest part of the systems i.e. on the hardware. Architecturally, the way they work is different from ESX, but to classify one as hosted and one as bare-metal is a little inaccurate. This is also quite a good introductory article to Hyper-V and it's architecture: http://technet.microsoft.com/en-us/magazine/2008.10.hyperv.aspx?pr=blog
Hope that helps,
Matt
#1
Matt McSpirit
(Homepage)
on
2009-04-11 21:07
(Reply)
Why would you record a video of yourself and put in on the Internet if you don't understand the material? This description is far from accurate. Is this person technical or marketing?
#2
what?
on
2009-04-12 03:45
(Reply)
While the original video is wrong in calling HyperV/Xen a hosted hypervisor, he is correct that there is a difference between them and ESX - I do not consider HyperV or Xen to be bare-metal hypervisors. They both require a "parent-partition" of some type, and whatever MS or Intel says, I'm 100% sure that there is some overhead in doing it that way. I checked out the Technet article linked above (btw, never trust a vendor's marketing about their own product), and they have a nice graphic showing HyperV sitting "below" the parent-partition - I don't believe it, HyperV sits "in" the parent-partition, and it uses the parent-partitions driver stack to access the physical hardware. This is where they get their advantage of a much larger HCL - the parent-partition already has the large HCL. And if all the IO and such is going through the parent-partiton, then I'm sure that it is at least occasionally having to wait its turn on IO that the parent-partition is doing (basically, I believe the going-shopping-with-kids analogy is correct)
#3
Rob B
on
2009-04-13 19:53
(Reply)
If you read Matt's linked article, the architecture is clear. The Hypervisor gets direct access to memory and CPU, however for all I/O related activity such as disk, it passes that to the root partition so it can utilize its drivers.
So it's kind of half-and-half, some is on the bare metal while the disk I/O gets bumped to the root partition which is touching the bare metal in those regards. Unfortunately that doesn't look as pretty in a marketing diagram so they fudge (e.g. lie) a little.
#3.1
JGrote
on
2009-04-14 04:17
(Reply)
haha maybe this guy should stick to buying christmas presents for his wife....
#4
Daniel
on
2009-04-13 22:16
(Reply)
