Convergence time for GLBP

[moabdallah]

Two Cat-6500 backbone Switches, each switch with single SUP720-3B supervisor

GLBP for loadsharing and reduandancy

Incase of any supervisor failed ( falied switch), What is the convergance time between these two backbone switches ?

Regards

Mohamed

[Roberto Salazar]

Routers on which timer values are not configured can learn timer values from the active virtual gateway (AVG). The timers configured on the AVG always override any other timer settings. All routers in a GLBP group should use the same timer values. If a GLBP gateway sends a hello message, the information should be considered valid for one holdtime. Normally, holdtime is greater than three times the value of hello time, (holdtime > 3 * hellotime).

http://www.cisco.com/en/US/products/ps6550/products_white_paper09186a00801541c8.shtml

For 6500 with sup redundancy look over the NSF with SSO:

http://www.cisco.com/univercd/cc/td/doc/product/lan/cat6000/122sx/swcg/nsfsso.htm#wp1023627

[moabdallah]

Hi,

Thanks for your reply

What is the most recommended protocol to achieve redundancy (min. convergence time) between two Cat-6500 backbone?

Is it HSRP, VRRP, OR GLBP? Any other suggestion

Regards

Mohamed

[Darren Ramsey]

I run GLBP timers of 1 hello and 3 dead on some subnets. On VOIP subnets I tune down to 250ms hello and 750ms dead. Tune your IGP down as well and less then 1 second failover should be possible.

[moabdallah]

The netwrok is running data and IP telephony traffic

Is it ok to use the GLBP on data & IP telephony network ?

Will I get session disconnect on IP Phones during failover between the two backbone ?

Regards

Mohamed

[devang_etcom]

when ever you are looking for the redundancy then HSRP and you want to achive redundancy as well as loadbalancing then GLBP...

GLBP is also best for the redundancy...so according to me GLBP is good...

regards

Devang

[kamlesh.sharma]

depend on your IGP timers how soon they converge. but in LAN side redundency prospective it's tunning is doable. If ur both the swtiches are connected to core with two link i mean one link to 1st core other link to other core then for outside network your convergence will not affect the network but for GLBP convergence it could be 1 sec and lesser than that.

but ultimately ur session is going to discunnect at the time of failiour becos application are so sestive to delay.

so my suggesion is to tune the GLBP OR VRRP OR HSRP and connect your distribution switches with both the core and you are on the fly that's all.

HTH