core switch redundancy

[mateens]

Hi,

school with around 800 users having one core switch 6509-E sup-720 (inter-vlan routing) collapsed core design connected to - 30 layer 3 HP switches with 10G and 1G backup links - 2 juniper WLCs 120 APs and VMware servers looking for a solution to achieve core redundancy. suggestions required,

[paul driver]

Hello

I guess it depends on your budget, your cheapest solution would be to have dual supervisors in the same chassis running SSO for resiliency and NSF is you have dynamic routing.

Review here

res

Paul

[Mark Malone]

Hi
Does the core have 2 sups that would provide some level of redundancy once it had to power supplys with separate feeds that would just leave the chassis as only point of failure and they rarely go but probably should have 2 cores really using HSRP or GBLP between them or VSS mode for proper resiliency/redundancy

If the HP switches support VRRP you could use one of them connected to the core on vrrp but it would need to be able to take the network load if it failed over from the 6500

wlc then to each core switch if it was there and dual link every server in port-channels and team the nics on their end

if only 1 core with 2 sups you could still dual link but to different modules on the 65

Thats just some options but personally i would have like for like core switches for a proper design , do you have budget for this or is it just work with what you've got

[mateens]

I am just preparing the best solution(s) for the school in the long run (5-10 years). Budget is not in my control but i think as it is an upgrade after many years we´ll manage the budget.

Spending money on 10 year old switch is not my first option and it might be soon Eol. so no another sup.

I am thinking of placing the switches in two different server rooms. But what requirements should i look for so that it would not be an overkill ?

We have HP2920AL48G mostly. Looking for some simple configuration. And school can manage a few minutes downtime does not have to be a millisecond failover but it should be without any intervention.

[Mark Malone · Accepted solution]

Hi

If it was me i would look for something VSS capable , that way each of the HP switches could be connected to each of the switches while still in a port-channel , even if you lost a full core switchthe  access layer would stay up , no need for re convergence etc as the VSS acts as one logical unit

if budget was problem another option is get a couple of high powered dist switch mmaybe 10gb 3850s and stack them together and link your access switch to each one again as its logical unit in a stack no convergence required if one fails , but VSS would be more resilient and powerful than that

again another core option not an ideal design as there DC switches but couple pf the newer cheaper 9k l3 switches that support VPC mode , there very powerful in hardware but not feature rich but if its not  complicated l3 setup would work well and ther not that expensive , again each access switch could be connected to each 9k in port-channel mode

The screenshots below show what each would look like physically , they would all give resilient options

For our office here i use VSS in my campus 2 x 6509-E and we only have about 500 users , thats purely though for access users really as  we have a DC for servers etc so i dont think any of those would be overkill as you have 800 users thats quite a bit of traffic

Aswell there is no set date yet even off Cisco for the 6509-E going EOL , there is newer switches available too like the campus 9ks 9300/9400s too

Cisco Catalyst 6509-E Switch

Specifications Overview

---

Series:	Cisco Catalyst 6500 Series Switches
Product ID:

Status:	Orderable How to Buy
Compatibility:

End-of-Sale Date:	None Announced
End-of-Support Date:	None Announced

[Leo Laohoo]

6500, nowadays, with Sup 720 is very expensive to run.  In some cases, because this model is already in the EoS table, the price of the maintenance contract has increased.  

With only 800 users, this design is an overkill.  

Once the Catalyst 9K firmware is stable, this would be the way to go. 

However, with current financial environment, I would go with a stacked 3850 (Enterprise Services) as a core or collapsed core/distro.  

[mateens]

3850 stack would be bit complicated config. ?

[Mark Malone]

Actually once the switches have the same model/software they will form a stack without any config and become one logical switch , you can manually set your master and priorities but its optional
when there stacked , say 2 switches together when you log in it will have ports 1/0/1 - 48 and also ports 2/0/1 - 48 as well and so on depending how many switches in the stack

[mateens]

and if budget is not a problem what would you suggest ?

[Mark Malone]

i would go with some form of core switch maybe not 65 bit overkill but id still get at least dual 4507-E sup8 in VSS mode , good 3850 stack would work just as well you can even get 10gb ones , but they both offer same form of protection n resiliency and redundancy, new 9ks as Leo said are the latest option we decided not to use them recently as there too new so we went dual 45s for now

[mateens]

Thankyou guys, I think 2 4500s with VSS would be first priority. There is one more problem though..... I cannot just go and buy these ,  have to prepare a document with requirements and make it available for bidding etc... 

Not sure what port density, uplink port speeds, features, protocols etc should i mention keeping in mind future growth, IoT, Cloud etc ? 

[Mark Malone]

This is just my opinion on it , but i would have at least a 10gb card in each switch for the up/downlinks to the 10gb HP switches to match the lan if possible as its already 10gb and that will future proof the speed for long time , you could even run port-channels down to the access forextra throughput , other cards could be 1gb
you probably need ipbase license for some l3 functions protocols , intervlan routing etc ,

The three feature sets available with Cisco Catalyst 4500E Series Switches are
:
●
Cisco IOS XE Software LAN Base: This feature set on the bundles provides Layer 2 features for access.
●
Cisco IOS XE Software IP Base: This featu
re set is upgradable with a software activation license (SAL).
It
includes all Layer 2 features and some basic Layer 3 features. ISSU and SSO are supported in this
package.
●
Cisco IOS XE Software Enterprise Services: This feature set is upgradable with a SA
L; it supports full Layer
3 protocols and advanced features such as complete routing scalability, Border Gateway Protocol (BGP),
Virtual Route Forwarding (VRF), Policy
-
Based Routing (PBR), and so on.

This link and the one below it show you the options you would have when purchasing a 4500e model , whats supported speeds and modules power supplys etc you can put together for the bill of materials for the vendor , also included a 3850 link if the pricing gets out of control on the 45s you can review wht models are available in that set

https://www.cisco.com/c/en/us/products/collateral/switches/catalyst-4500-series-switches/product_data_sheet0900aecd801792b1.pdf

https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/catalyst-4500-series-line-cards/product_data_sheet0900aecd802109ea.html

https://www.cisco.com/c/en/us/products/switches/catalyst-3850-series-switches/index.html

for now you shouldn't need to worry about anything cloud unless you have services , devices out there on something like AWS and even then swapping core switch shouldn't really cause an issue as everything of that nature would be coming over your internet switch/connection anyway , if you went vss the 2 45s would be just as capable as a standalone 6509e

[Leo Laohoo]

---

@mateens wrote:
3850 stack would be bit complicated config. ?

---

If you're doing static VLAN assignment then 3850 isn't very complicated.  

• IF going with 4500, make sure the chassis is 4500R+E and nothing else.  If it isn't "R+E" then the vendor is trying to peddle EoS stuff which may not be compatible with some supervisor cards and line cards. 
• Don't be tempted immediately with the Sup9E.  
• Don't be tempted to do dual-supervisor + VSS.  I'm a big user of 4500R+E and all I can say is the supervisor cards and line cards are very reliable.  I RMA more 2K & 3K but I've never RMA-ed anything from the 4K business line.
• Don't skimp on the power supply.  Get both of them.  
• Last but not the least, if you're unsure.  ASK. 

[mateens]

"If the HP switches support VRRP you could use one of them connected to the core on vrrp but it would need to be able to take the network load if it failed over from the 6500"

Do you guys see any downsides using this option ? 

[Mark Malone]

you wouldn't get the throughput that you would in a stack or a vss setup as they act as 1 logical core unit , its the least ideal setup but would offer some form of hardware redundancy at the core and was generally a standard design before VSS/VPC and stacks appeared offering more resilient setups , it would be definitely something you should test in a maintenance window if you went that way to make sure the HP could take the load the 6500 is already dealing with and that the failover is what you expect