11

We want to find a "safe" way to stretch layer 2 network across 2 data-centers, and since we don't have any equipment running NX-OS on either data-centers we can't use OTV. So what are the alternatives for us who have the following equipment at hand:

Data-center 1: 2x 6509-E in VSS (SUP720, IP Services)

Data-center 2: 2x 4500-X in VSS (IP Base at the moment)

8

VPLS, AToM, and L2TP are also additional ways to glue ports together that land on distant routers. Some of them require MPLS on all the in-between routers, but L2TP(v3 in particular) does not require anything besides ip routing on the intermediary routers. Basically, they glue together two remote router ports, and can usually pass spanning tree, etc, since they're not acting as switches in this regard.

| improve this answer | |
  • L2TPv3 is the poor man's MPLS (and I mean that in a good way). Plenty devices do it in hardware, too, but it doesn't look like the poster's are among those... – Niels May 10 '13 at 0:13
  • 1
    I don't think any of them work on his gear (4500-X) - would love to see them on 4500 though ;) – ioshints May 10 '13 at 13:58
  • Yeah, they are only support on the 6500 (given the OPs kit list and even then I'd say barely with a SUP720). – jwbensley Feb 6 '16 at 19:26
23

Dark fiber and LAG across the two links. VPLS won't work on 4500.

Disable STP on the DCI link (and make sure nobody ever reconnects the cables the wrong way) or use MST with each DC as a region (you don't want STP events in one DC to spill over into the other).

In theory you could use Metro Ethernet services, but if they don't pass LACP and STP transparently, you're in a total mess (read: don't even try).

#Rant On

Also, please note that there is no safe way to stretch layer-2 networks. Layer-2 network is a single failure domain - you get a loop somewhere, everything gets affected. You can limit the damage by using storm control, but then you're killing the good packets with the bad.

Finally, think about routing and connectivity implications if the link between the two data centers fails completely.

#Rant off

| improve this answer | |
  • I was going to post an answer to this, but, ioshints beat me to with the same substance of what I was going to say. So he gets an upvote. Like policy-based-routing, if you think you need to stretch layer-2, take a step back and figure out what it is that you're really trying to accomplish at its core. You'll probably find you don't need to stretch layer-2. – Jeff McAdams May 9 '13 at 14:15
0

There are many L2VPN Solutions, base on the time and budget you want to put in the project.

  • Direct physical link, just pull a cable across two DCs. (easy, expensive base on distance)
  • Virtual link, ISP will provide this if the two DCs are two far away. (easy, expensive)
  • Software solutions like vmware NSX and Cisco OTV (complex, and expensive, but they can do more than just L2)
  • Hardware solutions, if you look at some UTMs/routers/firewalls you will fine L2 extension function. (less complex, less expensive)
  • Low end software or open source solution, like OpenVPN, RouteOS, EOIP etc. (less complex, cheap)
  • L2VPN services, the l2vpnaas is very interesting, but it does L2VPN only. (easy, cheap, limited function)
| improve this answer | |
  • For an up-to-date answer, VXLAN should be mentioned which has the benefit to really scale. – Zac67 Jun 5 '18 at 20:55
  • Agree, and Vmware NSX is the best implementation of VXLAN. – Darwin Zeng Jun 5 '18 at 22:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.