WAE612 Redundancy: sessions failover and balancing process

[Cristian Giusti]

Hi,

As Cisco System Integrator we have developed a redundant WAAS solution for an Italian financial company big remote branch. We are working on a third-party projet, so we need support to understand what happens to the active sessions in the event of a hardware fault of one and / or both WAE.

We also need to understand how and when there is a balance of sessions in normal operation, as I have noted 80% of the sessions on WAE612A1.

I attach a simple diagram for logical architecture and the following you can read the basic configuration of the pair of WAS-612-K9 used for the solution.

Thank you.

Reguards,

Cristian.

############################## WAE612A1 ########################################

primary-interface Standby 1
!
interface Standby 1
ip address 10.37.28.34 255.255.255.240
exit
!
interface GigabitEthernet 1/0
standby 1 priority 300
no autosense
bandwidth 100
full-duplex
exit
interface GigabitEthernet 2/0
standby 1 priority 200
no autosense
bandwidth 100
full-duplex
exit
!
ip default-gateway 10.37.28.33
!
no auto-register enable
!
ip name-server 172.21.5.1
ip name-server 172.21.5.2
!
ip route 192.168.159.242 255.255.255.255 10.37.28.45
ip route 192.168.159.243 255.255.255.255 10.37.28.46
!
bypass static any-client 172.22.30.30
bypass static any-client 172.22.30.31
!
wccp router-list 1 10.37.28.45 10.37.28.46
wccp tcp-promiscuous mask src-ip-mask 0xf00 dst-ip-mask 0x0
wccp tcp-promiscuous router-list-num 1 l2-redirect mask-assign
wccp version 2
!
egress-method negotiated-return intercept-method wccp
!
############################## WAE612A2 ########################################

primary-interface Standby 1
!
interface Standby 1
ip address 10.37.28.35 255.255.255.240
exit
!
interface GigabitEthernet 1/0
standby 1 priority 200
no autosense
bandwidth 100
full-duplex
exit
interface GigabitEthernet 2/0
standby 1 priority 300
no autosense
bandwidth 100
full-duplex
exit
!
ip default-gateway 10.37.28.33
!
no auto-register enable
!
ip name-server 172.21.5.1
ip name-server 172.21.5.2
!
ip route 192.168.159.244 255.255.255.255 10.37.28.45
ip route 192.168.159.245 255.255.255.255 10.37.28.46
!
bypass static any-client 172.22.30.30
bypass static any-client 172.22.30.31
!
wccp router-list 1 10.37.28.45 10.37.28.46
wccp tcp-promiscuous mask src-ip-mask 0xf00 dst-ip-mask 0x0
wccp tcp-promiscuous router-list-num 1 l2-redirect mask-assign
wccp version 2
!
egress-method negotiated-return intercept-method wccp

[Bhavin Yadav · Accepted solution]

Hi Cristian,

Answers to your questions:

what happens to the active sessions in the event of a hardware fault of one and / or both WAE: The active TCP sessions will get reset and the new connections will be put in pass thru mode.

We also need to understand how and when there is a balance of sessions in normal operation, as I have noted 80% of the sessions on WAE612A1.: This is a bit complex procedure. You are using WCCP with L2/Mask. WCCP is not a load balancing protocol, but rather it is a load sharing protocol. You will have to tweak the mask value parameter to load balance the traffic fairly. The mask value depends on the ip address schema in your network, number of WAEs, and other parameters.

If you contact your Cisco Sales Account team, they should be able to help you derive a good masking value for your setup.

Hope this helps.

Regards.

PS: Please mark this as Answered, if this answers your question.

[Cristian Giusti]

Hi Bhavin,

     I think that you are referring to the the set of hash address buckets. I will study it with our Cisco Sales Account Manager.

Thanks.

Regards.

[ktunugun · Accepted solution]

Hi,

Please see below my responses to your questions.

Failure of one WAE: As Bhavin said in his response all the sessions handled by this WAE will get reset. While these connections re-establish the bucket assignment will also get re-caclculated and the buckets will be assigned to the remaining WAE / WAEs. The new sessions will go via the remaining WAE / WAEs based on the re-calculated bucket assignment unless there is an overload on the remaining WAE/WAEs in case which these new session would go passthrough.

Failure of both the WAEs: All sessions handled by these WAEs will get reset and subsequent new connections would go normally just like without WAAS scenario.

Load Balancing: As Bhavin said it is more of a load sharing and is dependent on the mask value you choose. As per the configuration the mask value you have chosen is of 4 bits which provides 16 buckets. These 16 buckets are distributed across 2 WAEs in your scenario. The recommended practice is to use a smaller mask value when you have fewer WAEs. For ex use 1 bit value (0x1 which gives only 2 buckets) for 2 WAEs, 2 bit value for 3-4 WAEs and so on.

Also you need to look at your source IP Address space. What part of the IP Address changes frequently. Based on that a suitable mask value has to be configured.

Please refer to the below URL that may help you calculate a correct mask value for your environment.

https://www.myciscocommunity.com/docs/DOC-9521

Hope this helps.

Regards

Kiran.

[Cristian Giusti]

Hi Kiran,

     thank you very much for the clarity of the response. I'll give you my compliments: WCCP mask calculator tool is really a great job. It will become one of my development tools ;-).

Thank you all guy.