Question 531 - 156-315.81 discussion

SDF (Sticky Decision Function) and Flush and ACK are the two correction mechanisms that are available with ClusterXL under R81.20.According to the ClusterXL R81.20 Administration Guide1, correction mechanisms are methods that ClusterXL uses to prevent or recover from out-of-state situations, which occur when different Cluster Members have different information about the connections that they handle1.ClusterXL supports two types of correction mechanisms: SDF and Flush and ACK1.

SDF (Sticky Decision Function) is a mechanism that ensures that packets of the same connection are always handled by the same Cluster Member, regardless of the load balancing algorithm. SDF uses a hash table that maps each connection to a specific Cluster Member, based on the 5-tuple of source IP, destination IP, source port, destination port, and protocol.SDF prevents out-of-state situations by avoiding the switch of Cluster Members for existing connections1.

Flush and ACK is a mechanism that synchronizes the connection tables of different Cluster Members when an out-of-state situation is detected. Flush and ACK works as follows:

When a Cluster Member receives a packet that belongs to an unknown connection, it sends a Flush message to all other Cluster Members, asking them to delete the connection from their tables.

When a Cluster Member receives a Flush message, it checks if it has the connection in its table. If it does, it deletes the connection and sends an ACK message to the sender of the Flush message, indicating that it has performed the deletion.

When a Cluster Member receives an ACK message, it creates a new connection entry in its table for the packet that triggered the Flush message, and processes the packet normally.

If a Cluster Member does not receive any ACK message within a timeout period, it assumes that no other Cluster Member has the connection, and creates a new connection entry in its table for the packet that triggered the Flush message1.