BIG-IP LTM provides various load balancing methods Under Pool > New Pool screen option. Most load balancing methods will allow you to choose between member and node. You will see the difference in the further article.
Basically, you can load balance between either nodes or members. Node is just an IP address (10.240.102.50) of the server, so LTM doesn’t check for port while making load balancing decisions, whereas Member is IP address plus Port (10.240.102.50:80).
F5 BIG-IP LTM supports various load balancing methods and they are broadly categorized as Static and Dynamic.
Static methods do not consider Server Performance while making decisions to load balance between its pool members.
Round Robin – All requests are evenly distributed across its pool members, so the 1st request goes to 1st node, 2nd request to 2nd node and 3rd to 3rd node. The new 4th request is distributed to 1st node and that’s how Round Robin keeps forwarding requests in its usual fashion.
When to prefer Round Robin? In an environment where servers are roughly equal in processing speed and memory.
Ratio(member) / Ratio(node) – All requests are distributed based on the static weight value that you define.
When to prefer Ratio(member) and Ratio(node)? Ratio is extremely useful if you have some servers with powerful processor capabilities and want to direct maximum requests to this server.
Dynamic methods do consider the Server Performance while making decisions to load balance between its pool members. BIG-IP LTM monitors the server availability by monitoring its pool of nodes
Dynamic Ratio(member)/ Dynamic Ration(node) – Ratio weights of members or nodes are based on real-timeserver performance analysis. The ratio weights are system generated and the values are not static. The necessary performance monitor software such as WMI or SNMP agent must be installed on the nodes to implement dynamic ratio load balancing method.
Least connection(member) / Least connection(node) – The server with least number of connection at the moment will receive the new request. LTM distributes connections based on the connection count between LTM and the pool members or node. It doesn’t account the connections that the server may have with the other systems, for example server connection to database.
Fastest – The server that responds faster for Layer 7 requests will receive the new request. Note that it doesn’t use Ping (e.x., Ping on Port 80) to decide while forwarding the new requests, instead it uses Layer 7 requests.
Observed (member or node) – Similar to Ratio but the Ratio value is assigned by BIG-IP LTM based on the total number of Layer 4 connections currently active on pool members. A pool member with a lower than average connection count is assigned a ratio of 3. A pool member with a higher than average count is given ratio of 2. These ratios are dynamically reassigned by BIG-IP every second.
Predictive (member or node) – Similar to Observed but determines whether a node’s performance is currently improving or declining and assigns more aggressive ration values . A pool member with a lower then average connection count is assigned a ratio of 4. A pool member with a higher then average count is given ratio of 1.
Least Sessions – This method selects the server that currently has the least number of entries in persistence table. In order for this method to work, virtual server must refer a profile type that tracks persistence connections, such as source address affinity or universal source type. The least session method is incompatible with cookie persistence.