March 08, 2009

Update: Cisco IT and the Nexus Deployment


image

Here is a quick update from Sidney on the progress with the Nexus and unified fabric deployment in Cisco’s production data centers.  Beyond Sidney’s update, you can get more details on the project here.

Omar Sultan Posted by Omar Sultan at 09:35PM PST

Permalink, Comments (3), Trackbacks (0)

Tags: cisco it fcoe nexus

3 Comments

Jim Loges Mar 10, 2009

Omar,
We are creating our first set of VMware server farms which will require 10GB ethernet.  We currently run our other server farm on the VSS1440 switching platform.  I was considering using the Nexus 5010 w/storage option and a fabric extender for this new farm.  This would connect to a 10GB connection in the Catalyst 6509 via layer 2 or 3.  Does it make sense to use the Nexus in this fashion or wait a few years and replace the VSS1440 with a pair of Nexus 7000?

Omar Sultan Mar 10, 2009

Jim:

Thanks for your comment.  So, I am not sure I can give you a definitive answer based on the info provided—your account team would have better context for a more complete answer.

That being said, a couple of general thoughts.  First, if all you need is 10GbE, then your existing VSS set-up with 10GbE line cards is a good place to start.  The one caveat is if you are substantially increasing the number of 10GbE-attached servers, then you might want to look at either Nexus platforms to give you sufficient 10GbE density. 

If you are looking at a unified fabric in your server access layer, then the Nexus 5000 makes a lot of sense for the simplification it offers.  The solution offers a lot of cost benefits and is fully supported by VMware.  One point I wanted to clarify is that the Nexus 2148T provides only GbE server access and does not support FCoE.

Hope this helps—feel free to ping me directly if you wish.

Omar

D2 Mar 12, 2009

Hi kids,

Main issue is L2/L3 arch for cross-site going forward. Larger L2 domains to facilitate vMotion+Storage and full mesh. C(I|T)O’s have been sold on utility/mobility of VM’s and bare metal re-addressed/re-purposed on the fly across multiple data centres. L3 to the edge? Better? DCE? Common edge? Fibre-Channel? FCOE?

Our architectures are more around pods and injecting host routes? LAM? Load Balancers? IGP aware endpoints?

Essentially the unique addressing challenge remanins as a node represents itself to the front end service networks(IP) and backend storage/mgmt networks (FC[OE]/IPFC/IP)...

Post a comment

Join the conversation!

We encourage your comments, questions and suggestions. All comments are moderated and will appear as soon as they are approved by the moderator.

Please increase the validity of your comment by providing a valid first and last name. Spam, off-topic or offensive comments will not be posted.

Name:
Email:
URL:

Comments:

Notify me of follow-up comments?

Submit the word you see below:


Post a trackback

Ping this URL to post a trackback:
http://blogs.cisco.com/trackback/7061/XXBZeG1G/

More blog posts

Previous post:
Data Center Interconnect

Next post:
New Thoughts on Data Center Temperature Management

Recent posts:
November 2009 Archive