Flexlink: Proof of concept

Introduction

Proof of concept for VRF, PBR and Flexlink will be based on this diagram.

Cisco 871 has reached its EoL on June 2010, this router is designed for branch office with a built in 4 port switch. The built-in switch itself has only one layer 2 redundancy mechanism – STP 802.1D, there’s no flexlink nor is there RSTP 802.1w. With computers operating at higher speed, recovery time of 50s is simply intolerable and in some situation cause unrecoverable lost, business sector such as financial sector cannot tolerate 50s recovery time as every second counts within 50s huge amount of money could be lost due to dealers unable to get the latest market update to react appropriately.

871’s built-in switchports are not flexible, it could not be turned into routed port like MLS (example 3550) , the solution for this is to use flexlink.

Flexlink

Flexlink is a layer2 redundancy protocol, in some situation if customers want to turn off STP but still need redundancy flexlink can be the alternative. The lab here is to test its recovery time.

Flexlink creates a loop free layer2 connectivity while at the same time provides link redundancy, on the diagram above, fa0/1 is the primary link and fa0/3 is the backup link, should fa0/1 fails fa0/3 will preempt and take over the active link role.

Mac address will be transferred from fa0/1 to fa0/3, and fa0/3 will send mac address update through multicast (dummy multicast) to inform upstream switch that there’s a mac address update. A faster solution for this default flexlink behaviour is to enable mac address move update, mac address move update will send mac address update to upstream switchports, the upstream switchports must support mac address move update in order to work.

Supposed mac address move update is transmitted from fa0/3 to upsteam switchport fa0/1, fa0/1 will update its mac address table, the MMU is transmitted before the dummy multicast is transmitted.

The default preemption delay time is 35 seconds, this behaviour can be adjusted between 1s and 300s, once the primary link is up, the primary link will wait for a period equals to the preemption delay before taking over from backup interface. Objective for this lab is to test how fast is the recovery time as compared to STP.

 

Time taken for backup interface to take over when primary link fails

Backup link fa0/3 took over when primary link fa0/1 failed. Recovery was 110ms which was seamless to enduser

 

Downtime while primary link fa0/1 took over from backup link fa0/3

10s downtime when primary link fa0/1 recovered from failure. Preemption delay was at 10s

 

12s downtime when primary link fa0/1 recovered from failure. Preemption delay was at 1s

 

No downtime when primary link fa0/1 recovered from failure. Preemption delay at default 35s

 

Conclusion

1. The time taken for backup switchport to inherit primary switchport’s cam table and inform upstream switchport about mac address update through dummy multicast was between 90ms and 100ms.

2. The time taken for primary switchport to re-inherit backup switchport’s cam table and to inform upstream switchport about mac address update through dummy multicast was between 10s and 12s.

3. Note that upstream switchport’s cam table has to clear off and accept the latest cam table update, this takes time as well.

4. Note that only one host was used to test, if there’s multiple hosts sending the recovery time might increase.

5. It is best to leave preemption delay at its default 35s to give switches ample amount of time to clear and re-populate their CAM tables.

 

Configurations that enabled flexlink

interface FastEthernet0/1
switchport trunk encapsulation dot1q
switchport trunk native vlan 100
switchport trunk allowed vlan 10,20,100
switchport mode trunk
switchport backup interface Fa0/3
switchport backup interface Fa0/3 preemption mode forced
end

interface FastEthernet0/3
switchport trunk encapsulation dot1q
switchport trunk native vlan 100
switchport trunk allowed vlan 10,20,100
switchport mode trunk
end

Preemption delay timer not shown in configuration because it was at default 35s.

Advertisements
This entry was posted in High Availability, Switch and tagged , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s