Layer 2 vs Layer 3 Access Closets

[gdrandles]

I am trying to determine whether to configure L2 or L3 between the core switch and the access closets.  The core switches are 6509-E and the access closets have 3750x-48P-L switches (stacked).  There are 5 switches on the second floor and 3 on the first floor.  I have fiber running from the access closets and the core switches.  Performance wise is it better to create trunks or routed interfaces between the core switches and the access closets.  If I went with Layer 3, would EIGRP be preferred.  I am already running OSPF on the core switches as I have Juniper routers connecting via MPLS to our remote offices.

Thanks in advanced.

[Guohua Zhang]

I do not see any reason to run it over Layer 3. Do you have?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

There can be some advantages running L3 at the edge.

When you have dual uplinks to individual core or distribution routers, L3 at the edge makes it a bit easier to use both links actively.

In such a topology, L3 at the edge avoids some unicast flooding situations.

If you have multiple subnets at the edge, L3, of couse, will route between them locally without the need to bounce across the uplinks.

If an uplink fails, L3 might reroute around the failed link faster than L2 STP.

As to OSPF vs. EIGRP, both can work well.  Unlikely switching from OSPF to EIGRP provides much of an advantage, especially since you have brand J boxes.  However, if what you have in mind is using EIGRP stubs included in base image (?) and not licensing the full services image, there is a cost savings.  (NB: not sure about the X series, but in original series, legally, you only needed to license one advanced image [or two for redundancy] per stack.)

[Latchum Naidu]

Hi,

I would say a trunk port between core and access is fine and morever you are running fiber which is very good for speed fast.

Regarding your remote sites communication, I would suggest EIGRP is best. Even if you running with OSPF still you can use EIGRP and redistribute another into it.

Please rate the helpfull posts.
Regards,
Naidu.

[Rahul Kachalia · Accepted solution]

Many advantages :

- Simplicity - Single control-plane EIGRP or STP, instead full package of Layer 2 protocols to build multilayer topologies. No FHRP any more, means no need to tune sub-sec timers  on each SVI this results better CPU performance.

- Manageability - Just routing protocols. No STP Root, Root Guard, LoopGuard/BA, FHRP (HSRP/VRRP/GLBP), with stub routing function just default-route

- Reliability - No more concern on STP loops in dist-access block.

- Availability - Optimal, redundant and bi-directional fwding paths based on IGP, so no more asymmetric switching.  CEF based Equal-Cost Multi Path (ECMP) switching, network recovery on individual link failure is within ~200msec (VoIP baseline) rather protocol timer-driven between aggregation which could be several seconds (default) or ~1 sec with aggressive timers.

Refer to Chapter 2 for many additional details :

Borderless Campus 1.0 Design Guide

thanks,

rahul.

[Jon Marshall]

Just to add to the others.

They both work fine ie. L2 or L3 but the answer to your question really depends on how your network is setup ie.

if you go L3 from the access-layer switches the vlans on each switch (or pair of switches) are isolated from other vlans. What this means is you cannot have the same vlan on different switches. So if you needed a vlan on the one of the first floor switches and one of the 2nd floor switches, or you needed a vlan on 2 of the 2nd floor switches that weren't interconnected L3 will not allow you to do this. You would need to use L2 uplinks.

L3 does have it's advantages which the other posts have outlined but it is not as flexible as L2. Apart from the vlan issue above there is also service placement ie. firewalls/load-balancers etc. With L3 in the access-laye3r you cannot run any of these in transparent mode from the core. SPAN/RSPAN can also be a challenge.

So i would say in a campus environment L3 does work well but for a DC or server environment it is not necessarily the best especially with the introduction of VSS/VPC technology.

Jon