We have a new office that we need to connect with a point-to-point leased line. All providers are giving Ethernet terminated links. One of the links is a 100Mbps fiber line. The other is a 4Mbps link with Riverbed appliances at both ends. The 100Mbps line is the primary link.
100Mbps - fiber to Ethernet converter
4Mbps - serial (2 x 2Mbps multiplexed on a mux device supplied by provider) - converted to Ethernet
What model of Cisco router will enable us to achieve seamless failover between these two lines.
Both lines will be Ethernet terminated into the router.
What protocol should I use to achieve failover between the two links?
If I can also be pointed to a configuration example, I'd be very grateful.
Why does it need to be a router? You could use a layer 3 switch and get a better bang for your $$ or you could use an ASA if you need a firewall between sites. The switch you could simply run a routing protocol accross it, the ASA you could do object tracking or even a routing protocol.
Thanks for responding. Do you think you can throw a little more light on this? How would we achieve failover between the two links using an L3 switch? What protocol would we have to use?
Please dont top post, makes it hard to follow/reply to the threads. A layer
3 switch is simply just a big ole router, you don't need the ability to terminate T1's so a true router would just be a waste of $$. Any of the routing protocols will do load balancing and failover. It really depends on how the riverbends play. They won't support eigrp but they will do ospf. If the providor won't use a dynamic then you could build gre tunnels from lan to lan and run the protocol accross those.
the 2 links are layer 2 (ie the provider is not routing between the ends of the links you connect to). no need to firewall (ie these links dont hook into the Internet or another insecure network).
100M is better than 4 Mbps, so you want to use the 100M normally, and fail over to the 4 M when it breaks. both link terminate at the same place in the sites. no need for extermely fast failover - a few sec to 10s of sec is OK no need for any other resilience.
if you dont think you have or can find these answers, then go and find a reseller / installer / etc who can sort this for you
Note - getting someone with experience involved would be a good idea anyway, since it sounds like you dont have much background and this is an area where mistakes can be expensive.
So - design.
at each end of the link install a layer 3 switch. run a routing protocol on each box to protect each WAN link - i prefer OSPF, but just about any would do. bias the costs so when both WAN links are up traffic follows the 100M link. tweak the timers to improve failover time if the default isnt good enough.
Test it..... both for performance and failures, several times for every kind of breakage you can think of (Ideally get the carrier to drop the link in their cloud so you can check for the kinds of faults they expect).
Note - once you have a resilient network, you should have some way of monitoring if anything breaks since the "self healing" means breaking the back up is invisible to the users - often back up links seem unreliable because no one notices they are not working until you need them....
For L3 devices, minimum switch would be a Cisco Catalyst 3560, and you need a model with enhanced software to support a good routing protocol. List on the minimum spec for this is around $5k / unit.
2 would meet your stated requirement - but there is always something else.......
if you need more resilience then you need a designer to sort out the interactions since it will have to integrate well into your LAN at each end.
Yeggi, You are going to want to use dual Cisco 7201's. They are expensive, though I have 2 in a rack, and failover is seamless, uptime is unbelievable. The router has purely been designed for 100% uptime. Cheers, Alex
Email me with any more questions, snipped-for-privacy@turnerzworld.com
Layer 3 switches don't do QoS queuing, they do CoS queuing. They can also do QoS marking, but you can't traffic-shape, do WRED or any of the other fancy QoS queuing stuff with a Layer 3 switch. The exception is the Metro Series 3750's which have two ports that you can do Layer 3 queuing on.
Sorry - wrong. Cat 3560 / 3750s have DSCP mapping to queues.
You may not like being limited to 4 queues, and having to find somewhere for OSPF etc (which on a router get their own hidden queue) - but everything in the box can be set up to work via DSCP.
Now - you can set them to use 802.1p CoS, or the old IP ToS bits as well (and CoS might even be the default) - but QoS is in there......
They can also
Agreed - the special ports on the 3750 metro are much more flexible - you can traffic limit by VLAN and AFAIR by VLAN inside a stacked VLAN (work uses this for multi VRF traffic control on Ethernet WAN links for MPLS services).
But some of those features are on 3560 (and 3750 since the chipset & s/w is pretty much identical).
So, i think you are confusing a specific implementation on these boxes with a generalisation.
If you widen the discussion, then just about every permutation of WRED, Qing, shaping / rate limiting and so on is on a switch made by someone
I used to work on Nortel 8600s which have a big chunk of these features (but a different set of pain points). Foundry also make some nice stackable switches - and 8 Qs and proper working thresholds sounds much more useful / flexible......
Yes - 3560 / 3750 do support shaping (although only at multiples of 10% of line speed) and no they do not have WRED.
FWIW other Cisco switches have different tradeoffs - Cat 6k with Sup 720-b and DFCs does have WRED, but no shaping at the same time if you use 67xx line cards.
If you need an upmarket stackable then Cat 6524 looks interesting - but 2 to
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.