Hsrp/vrrp/glbp

notgoing2fail

Quick question guys, if 2 routers or 2 layer 3 switches are taking part of a redundancy protocol, but they are then connected to ONE switch which provides access to hosts. So I suppose you can call it the access layer switch, but if that switch fails, all the hosts lose access anyways!!

So am I missing the grander picture here? You go out of you way to create redundany or load balancing between 2 switches but then your central point of failure becomes the one access layer switch that connects the hosts...

burbankmarc

Yeah, pretty much.

Just use 2 access layer switches plugged into both layer 3 switches.

zerglings

For users, yes. For servers, you can connect the NICs to two different switches for redundancy.

When I was doing field work, I get more requests of letting the network down for more than the expected downtime than requests for having it up as soon as possible. c",)

notgoing2fail

burbankmarc wrote: »

Yeah, pretty much.

Just use 2 access layer switches plugged into both layer 3 switches.

So the hosts/servers would both need two NIC's each on each separate access layer switch....

zerglings wrote: »

For users, yes. For servers, you can connect the NICs to two different switches for redundancy.

When I was doing field work, I get more requests of letting the network down for more than the expected downtime than requests for having it up as soon as possible. c",)

Ok, I think you answered my question above. Of course the hosts probably don't normally have two NIC's but I can see most servers these days come with two NIC's...

zerglings

notgoing2fail wrote: »

Ok, I think you answered my question above. Of course the hosts probably don't normally have two NIC's but I can see most servers these days come with two NIC's...

I hope I did. c",)

ccie1yr

notgoing2fail wrote: »

So the hosts/servers would both need two NIC's each on each separate access layer switch....

Servers are recommended to have dual NIC and perform loadbalancing. While for users you can split them over to both switches so that only half of them get affected in case of any failure.

DevilWAH

yes server only need two nics each going too a seperate switch. if the switvhes are stacked togather then you can create and ether channel to loadd balance. or of course you may be using a 4500 / 6500. in which case each nic goes to a seperate blade. as long as the switch has redundent supervisor units and powersupplies you remove the singel point failer. in fact with the new 6500, you can even create ether channels across physical seperate switches. but this is only for things like servers. the idea is that your core is redundent. if you have 50 access switchs you don.t mine if one goes down as it only affects that area. if you sites core router fails and you dont set up redundeny then the entire site fails. in general redundency costs money so you start at the core and work out as far as money allows. I have seen compines that run duel nic / etherchannel to desktops. Where uptime was critical for monitering and control of machinery.