Junos OS

Hi.

I tried configure VPLS with two sites. One of them is multihoming.

I tested some different configuration then multihomed site connected to one PE and multihomed site connected to two PEs ( see attach).

I have some question:

1) Switching between links on multihoming site is ~ 0.5 second. It is bad =( Can i improve this time?

2) Recovery time from backup link to primary link or from  backup site-1 to primary site-1 is 30 second =( How i can decrease it time?

3) Anybody use multihoming VPLS in production?

Attachment(s)

VPLS-VSTP.vsd.pdf   132 KB 1 version

Hey,

Are wa talking about BGP or LDP based VPLS?

Would it be possible to share the configs (if not full, then interfaces, instances, protocols).

Thanks;

About BGP auto discovery.

Hey,

In that case things that come to my mind:

- the same Site-ID for instances connected to the same multihomed site;

- unique RD for all PEs involved;

- if L3 IRB interface is configured inside the instance, "connectivity-type ce" should be set;

- Verify the STP (MSTP/RSTP) situation between CEs and PEs - here's brief explanation: http://data.proidea.org.pl/plnog/5edycja/materialy/prezentacje/EmilGagala.pdf - Slide46;

Thanks,

1) Site-id is same.

2) Unique RD on all Sites

3) No IRB

4) Cool link, tnx. But nothing changed

If i used VSTP and setup minimal timers for VSTP/RSTP switching between multihomed site is 6 second.

If i totaly disabled VSTP  - 120 second.

If i connected multihoming l2-domain to ONE PE and configure "active-interface any" the repairing time is 0,6 second

For my configuration i use that manual http://www.juniper.net/techpubs/en_US/junos10.0/information-products/topic-collections/feature-guide/vpls-multihoming-bgp-signaling-solutions.html 

Resolve my problem with tunning spanning-tree:

1) Configuring ports on cisco switch to p2p (By default it is "Shr" state)

spanning-tree link-mode point-to-point

 2) Configuring minimal "forward time". 4 sec

After 1) and 2) steps i had subsecond recovery time.