Installing Cisco 6500 virtual switching

[spidey07]

This technology enabled by yet another 6500 supervisor allows two 6500s to act as one. Enabling the holy grail of high availability - link aggregation/etherchannel across multiple chassis.

No more spanning-tree, no more complexity, just easy blocks of 6500s that appear as a single switch. Pretty cool actually.

It's worked pretty well in the lab and now it's hitting a new data center. The main gotcha is a "dual-active" (both supervisors think they own control plane) which could cause an unrecoverable loop. But there are ways to detect that. There's some network voodoo under the hood to allow this to work but the main thing is you can do very high available networks AND use all capacity of the switches.

Biggest think is link aggregation to servers - one or two connections to one switch, one or two to another switch, all links being used at layer2 and up.

required reading

 

[RadiclDreamer]

Very nice, we are building a new DC where I work and are looking at about 5 of these puppies. Might have to look into this further

 

[spidey07]

Originally posted by: RadiclDreamer
Very nice, we are building a new DC where I work and are looking at about 5 of these puppies. Might have to look into this further

You should. It depends on what architecture you need.

With blade servers being the "new thing" and "virtualization" this VSS feature can help.

I've said it for 6+ years, we're going back to the mainframe. All we're doing is spreading the different parts around like mainframe technology has done so well. Yet reinventing the wheel seems like a good idea.

 

[multiband8303]

Originally posted by: spidey07

Originally posted by: RadiclDreamer
Very nice, we are building a new DC where I work and are looking at about 5 of these puppies. Might have to look into this further

You should. It depends on what architecture you need.

With blade servers being the "new thing" and "virtualization" this VSS feature can help.

I've said it for 6+ years, we're going back to the mainframe. All we're doing is spreading the different parts around like mainframe technology has done so well. Yet reinventing the wheel seems like a good idea.

Tends to be every 2-3 years we have the cyclical "consolidation vs distributed" computing solutions.

Right now we are at consolidation, from an infrastructure standpoint, I agree we are going mainframe. With cloud computing however it will be a mixed bag of the infrastructure, network being consolidated with the new "VIRTUAL" switching, however application virtualization going towards the distributed route.

 

[tranceport]

At VMWorld 2008 the big thing was cloud computing. At Cisco networkers it was Nexus.

I'm about to upgrade the switching in 2 datacenters. Not sure I can justify these though. Currently looking at 4500 series gear. The 6500's just arn't really needed.

Plus with SupV on 4500's you can do GEC/FEC across multiple chassis now as well. This is really the only feature we need. Have Sup2+ right now.

 

[Genx87]

Originally posted by: spidey07

Originally posted by: RadiclDreamer
Very nice, we are building a new DC where I work and are looking at about 5 of these puppies. Might have to look into this further

You should. It depends on what architecture you need.

With blade servers being the "new thing" and "virtualization" this VSS feature can help.

I've said it for 6+ years, we're going back to the mainframe. All we're doing is spreading the different parts around like mainframe technology has done so well. Yet reinventing the wheel seems like a good idea.

Yup, we are looking at consolidation of our servers using VMware. When we complete that we are going VDI for anybody with a desktop. It just makes so much more sense than buying new desktops and having to deal with the migration and installation of applications and the hardware support. Give them a thin client and have them rdp into their virtual desktop. The only time they will change is when that OS is officially not supported by MS. If you use Vista that could be 10+ years of not having to replace a users desktop.

 

[drebo]

This is definitely one of the coolest things I've seen in a long time.

Can't wait til I have an opportunity to work with it.

 

[Gand1]

Oooh nice! I'm in the money hunt to replace our current 4000's with 6500 since we have outgrown them... Another good selling point! Thank you!

 

[spidey07]

In testing we're seeing <500 ms failover without any special tuning. A few things to watch out for:

1) make sure you enable NSO on your routing protocols
2) Use PAGP+ for dual-active scenario. Any etherchannel to another switch you're using this on will need to be PAGP+ aware - 3750s, 4500s, 6500s depending on software version
3) PFC and DFC operating modes. If you have a mix of DFC3c and DFC3cxl you'll have to tell the switch to operate in PFC3c/DFC3c mode, how to do is in the doc I listed.

 

[jlazzaro]

Originally posted by: Gand1
Oooh nice! I'm in the money hunt to replace our current 4000's with 6500 since we have outgrown them... Another good selling point! Thank you!

i wouldn't call VSS a definate selling point to 6500's...unless you're in a datacenter environment with the need to forward terabytes a second, the cost will surely outweigh the benefits.

it's just a shame Cisco took so long to deploy this technology, and really only made it available to the extremely high-end. what about those customers who cant stand the RSTP failover times but don't want to manage an entirely routed L3 core? an SMLT variant of some sort would be nice, and way overdue.

 

[spidey07]

Originally posted by: jlazzaro

Originally posted by: Gand1
Oooh nice! I'm in the money hunt to replace our current 4000's with 6500 since we have outgrown them... Another good selling point! Thank you!

i wouldn't call VSS a definate selling point to 6500's...unless you're in a datacenter environment with the need to forward terabytes a second, the cost will surely outweigh the benefits.

it's just a shame Cisco took so long to deploy this technology, and really only made it available to the extremely high-end. what about those customers who cant stand the RSTP failover times but don't want to manage an entirely routed L3 core? an SMLT variant of some sort would be nice...

meh, if you have the need for a lot of 1000 Base-T ports the price per port is actually low compared to other platforms especially when you add 10 gig.

 

[Cooky]

Does it support redundant ACE modules?

 

[spidey07]

Originally posted by: Cooky
Does it support redundant ACE modules?

Next release SXI, should be out within a month or two.

Although you'll have to check my accuracy as ***I believe*** only the network analysis and possibly firewall module were supported with SXH release, very limited service module support with SXH (current).

I'm chomping at the bit for ACE and WISM support.

Did the install today with 6 chassis, very easy to do if you follow the docs. The PFC/DFC modes caught me as I thought they would. All DFCs were 3c, the 10 gig cards ship with 3cxl.