Load Balance through ISP EPL service

[localhost123]

Hi,

thanks in advance for anyone that can help on this.

Consider this scenario:

Site A: 5406 connected to 2 differnet L2 Fibres EPL provided by ISP

|

ISP Switches (each fibre ends into separate switch we don't manage)

|

Site B: 5406 connected to 2 differnet L2 Fibres EPL provided by ISP (point to point from the above)

question is: how can i load balance between the 2 fibres since the devices in the middle are not configured for LACP or trunking?

Option 1: MSTP and each link on a different VLAN. This could be, but the servers VLAN is only one so i end up using the same link all the time even with MSTP.

Option 2: Create a trunk between the two ISP fibre ports on both sita A and B 5406s  (even if the ISP side switches will not have trunk enabled), and enabled Layer 3 trunking globally.

My concern is this not going to work because the middle switches dont have trunking enabled, but since ISP only managed L2 if i turn on trunking on L3, this might actually work.

Thoughts?

thanks

[parnassus]

I think Option 2 is not a viable option because, generally speaking, Port Trunking (no matter if then you set it to use LACP/No-Protocol) doesn't allow intervening devices or non-trunking devices...so if there are two switches ([*] Are really two Fiber Optic Switches or are two Media Gateways that collect your ISP incoming links?) you don't manage (and that your ISP don't want/can't configure accordingly) that aren't yet set up to cope with your HPE 5406's Trunk settings...it's going to be really problematic.

[*] Probably Port Trunking can work if links between your two Switches are transparently managed by two Media Converters (let's say F.O. <--> Copper or vice-versa)...these devices will add latency...but should be transparent enough to Layer 2 protocols.

[localhost123]

thanks heaps. Mmm. I guess i will try and see what happens. ISP tech has been a bit vague. They say they simulate a point to point but they have Cisco gear so ....

[Stephen A Swain]

You need better advice from the ISP's, they should be able to say what is and what is not supported as a technical level.

I'd avoid layer 2 redundancy myself, including trunks, lacp, and/or STP. You need some way to determine connectivity and failover, and the underlying switches and tunnels on the ISP side may cause problems here (are multicast or broadcast filtered?).

For example, a "native trunk" link depends upon interface up/down, and may not know if the ISP link is down somewhere else.

Another example, LACP depends upon BPDU's that may not get across the link.

STP will probably has issues with BPDU's as well, and will end up wasting a link.

What about OSPF and ECMP, or BGP and private AS's?

It would enable layer 3 connectivity monitoring, and load balancing.

But yeah, the ISP techs should be able to give you an indication of what protocols are supported on the client side. They said they simulate a point-to-point on Cisco, but they should have some idea of what their clients are doing generally, and be able to give you a shortlist of what is good and what isn't, that is if you speak to the right person.

This is kind of a WAN case really, my preference would be to use WAN gear, like the Cisco ASR series, better toolsets for the job.

If you require layer 2 end to end and connectivity monitoring, you may need something better than a switch, heading into the IWAN areas of products.