Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. If you'd prefer to explore, try our test area to get started. And see here for current known issues.

New Member

VM-FEX

Hello,

After having seen the VMmarks 2.1.1

http://www.vmware.com/a/vmmark/core/32/?s_score=d

And checking that UCS VMmark reaches the score using VM-FEX (http://www.cisco.com/en/US/prod/collateral/ps10265/vmm_infrastructure.pdf)

One comes to the big question:

Taking into account that score is almost the same as HP score (11.30@10 tiles vs 11.13@10 tiles), question is ¿where's the almost 50% supposed improvement in CPU performance one can see in the following link?

http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/ns944/white_paper_c11-593280.html)

Thanks,

  • Unified Computing
Everyone's tags (1)
8 REPLIES
New Member

VM-FEX

No insights on this?

Cisco Employee

VM-FEX

Dani,

It says when using Hypervisor Bypass, the VIC operates nearly 50% faster than using the standard hypervisor vSwitch.  I don't believe it's referring to it being 50% faster than HP...

Robert

New Member

VM-FEX

Robert,

As HP doesn't have VM-FEX, being faster than using the standard vSwitch and being faster than HP should equal the same...shouldn't it?

Cisco Employee

VM-FEX

They aren't testing against the platform (HP/Dell/IBM) in this whitepaper, they're testing the Cisco VIC on the vSS vs. VM-FEX.  This is the only way to compare apples:apples.

Each HP platform as multiple NIC offerings - we would declair an exact measured performance increase against them all.

Robert

New Member

VM-FEX

I agree that the white paper is not a test against HP...but you should agree with me that VM-FEX is a UNIQUE solution in the market as for performance (or at least that's what Cisco marketing says), and so HP/IBM/DELL can offer a myriad of NIC offerings, but none of them consists of a Hypervior Bypass (VMDirectPath). Therefore all of them equal to a standard vSwitch solution as for performance...that's why I'm trying to extrapolate the test results of the white paper.

In what aspect am I wrong?

New Member

VM-FEX

Hello Dani,

VMDirectPath is supported by all server vendors in some form or another.  The competitive difference for the UCS is the ease with which it is implemented.

* UCS looks just like a regular DVS and offers QoS capabilities

* No need to install special device drivers in the VM as you would for other solutions.

* Until vSphere 5, VM-FEX was only solution that allowed vMotion with VMDirectPath

etc. (I'm not in marketing, so I cannot rattle off the complete list of benefits)

So, we need to be careful with our messaging so that we don't give HP an opportunity to accuse us of fabricating info.  Let's have them publish all the fairty tales like the Tolly report.

Regards,

Trevor

======================================================================================

If my answers have been helpful in any way, please rate accordingly.  Thanks!

====================================================================================== If my answers have been helpful in any way, please rate accordingly. Check out my blog for useful vSphere & UCS tips: www.vmtrooper.com
New Member

VM-FEX

Trevor, VM-FEX allows VMDirectPath with vMotion SINCE vSphere 5 not UNTIL vSphere 5...plus it IS the only solution in the market that can do so (not WAS) as far as I've understood...maybe I'm rong at this point...

And as for performance, are you telling me that HP (11.13 @ 10 tiles), Fujitsu (11.02 @ 10 tiles), Dell (10.49 @ 10 tiles)

and IBM (10.29 @ 10 tiles) are all of them using VMDirectPath in their VMmark tests?? I don't believe so...

New Member

VM-FEX

Thank you for the correction.  I was misinformed by the staff who told me otherwise.

Regarding how the competitors arrived at their numbers, I have no idea of their configurations and whether they used VMDirectPath or not to get their marks.

I will echo Robert's earlier caution and advise against extrapolating competitive results based on a Cisco-only test White Paper.  It would be best to consult Cisco's Marketing\PM teams regarding what our competitors can\cannot do.

Regards,

Trevor

======================================================================================

If my answers have been helpful in any way, please rate accordingly.  Thanks!

====================================================================================== If my answers have been helpful in any way, please rate accordingly. Check out my blog for useful vSphere & UCS tips: www.vmtrooper.com
640
Views
0
Helpful
8
Replies
This widget could not be displayed.