Stickness configuration check-up and ...

Unanswered Question
May 2nd, 2010

I would like to check the following stickness configuration on ACE module.

first of all, I created the class-map vip statement as following:


class-map match-all EDM_DRM_TO_WAS_VIP

match virtual-address a.a.a.a tcp eq 8081


and next, I created the loadbalance statement as..


policy-map type loadbalance first-match EDM_DRM_TO_WAS_Traffic

class class-default

  sticky-serverfarm SRC_IP_STICKY


and I made a stick configuration as following:

sticky ip-netmask address source SRC_IP_STICKY

timeout 10

replicate sticky

serverfarm EDM_DRM_TO_WAS_FARM

< loadbalance policy>

policy-map multi-match EDM_DRM_TO_WAS_POLICY


  loadbalance vip inservice

  loadbalance policy EDM_DRM_TO_WAS_Traffic

  loadbalance vip icmp-reply

< Deplaying a Stickness in Server Vlan >

int vlan 20

desc ## Server Vlan ##

service-policy input EDM_DRM_TO_WAS_POLICY

Is it right for applying in server vlan ?

What is the main difference between hash in serverfarm and stickness?

I think that It is same function, but Why should I use a stickness configuration?.

Could you explain it more detailly?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Gilles Dufour Mon, 05/03/2010 - 06:21

The hash function is a mechanism that takes the source ip address, and the number of servers and gives an index which represent the

server to use.

The same ip, will always result in the same index.

Therefore it is considered some kind of stickyness mechanism.

However, if you change the number of the servers or if you re-order the servers, the resulting hash will be different and your client will then be send to a different server.

On the other hand, stickyness works in combination with a predictor algorithm.

The predictor determines the server to use for the first time the user is detected.

Once the predictor has selected the server (roundrobin, or leastconn, or ....), the sticky functions saves the result in a sticky table and each time the use comes back, we use the cached information instead of calling the predictor again.

If you change the predictor, add or remove servers, or reorganise servers, that does not affect the cached entry.

So your user will continue to hit the same server.

Before using sitckyness, you should verify you have allocated enough sticky resource.

Don't forget sticky resource are not like the others.

The value you allocated (even if called minimum) is the maximum the device will use.



This Discussion