2

enter image description here

I have a tech startup environment where I'm replacing consumer-grade routers with enterprise-grade.

Now, they are asking if we can have dual ISPs to minimize downtime and any disruption.

I have:

  • VLAN 70: 10.70.0.0/16 with 10.70.0.1 default gw
  • VLAN 5: 10.2.10.0/24 with 10.2.10.1 default gw
  • Mgmt VLAN is 70 where router has 10.70.0.2, layer 3 switch has 10.70.0.3, and so forth.
  • Virtual-IP GW of 10.70.0.1 or iBGP?

Usually, I use GLBP with two Cisco routers for CEF load-balancing, but we are not sticking with Cisco this time. Shall I use VRRP between vendors or use BGP with our ISP gateway being iBGP neighbors with each other?

Requirements:

  1. Later on, we want hybrid cloud with AWS and it uses BGP that I'm eventually going to set a default route in OSPF for.
  2. Minimize CPU on routers
  3. We are very ARP Layer 2 loop prone due to multiple adapters on servers acting unexpectedly.

In the long run, which would provide simple yet effective HA and load balancing?

  • Is this for hosting services or for internet access? Or both? – Teun Vink Nov 4 '15 at 12:21
  • Hosting 0 services(except VPN) but ONLY internet access – opensourceworker Nov 4 '15 at 12:21
  • A drawing would be nice. Is there only one layer-3 switch? Also, VRRP doesn't load balance; like HSRP is chooses one router at a time as the default gateway. – Ron Maupin Nov 4 '15 at 14:57
  • I think it's GLBP (Cisco proprietary) that does it and got confused, thanks. I'll have to use dual Cisco ISP GW then. But to get back to original question, Using GLBP/CEF on top of AWS BGP or just use BGP for ISP redundancy as well? Let me get on Visio but basically one router connected to each ISP router to main L3 switch. – opensourceworker Nov 4 '15 at 15:30
  • Drawing/Image added – opensourceworker Nov 4 '15 at 15:38
1

If you have a single layer-3 switch to which the LAN connects, you don't need an FHRP for the LAN. The question then becomes how to load balance from the layer-3 switch to the Internet.

If you need the layer-3 switch to know about the specific routes advertised from each ISP in order to make intelligent decisions about which direction to route particular packets, then BGP would certainly be the right choice.

Any FHRP will need to have a layer-2 connection between the routers. That would make the uplinks from the layer-3 switch be layer-2 links, but that doesn't seem to be how the layer-3 switch should be used.

I really don't see the need for two separate routers on each ISP link. That could be handled by a single router.

Another way to go is to have the links from the layer-3 switches toward the Internet be routed links, and just use the default route on each one. The default routes could be equal cost, and, depending on your layer-3 switch, could naturally load balance.

You should think carefully about load balancing. You can end up with asymmetric traffic flows which may cause problems. You can also slow down because of the out of order packet delivery problem. There is no guarantee that to and from a particular place on the Internet is equidistant from your site on each of the ISPs, or even that to a place and from a place will use the same ISP (asymmetric).

  • thank you for your long detailed answer. I'll have to study and design a new approach to avoid assymetric route. didn't think about that. thanks again – opensourceworker Nov 4 '15 at 16:55
  • maybe I'll use PBR – opensourceworker Nov 4 '15 at 16:59

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.