Server load balancing strategy

1. Static load balancing algorithm

Polling (RoundRobin): a request to cycle to cycle each connection server, wherein when a server fails 2-7 layers, out of the circular queue sequentially, a polling without participation of their normal .

Ratio (Ratio): assign a weighted value to each of the server to the ratio, according to this ratio, the user requests assigned to each server. When a server fails wherein 2-7 layers, out of the circular queue sequentially, a polling without participation until it returned to normal.

Priority (priority): to the server all packets, for each set of custom priority, assigned to The BIG-IP user requests a high priority group server (or polling rate in the same group, the user's request allocation), when All server fails highest priority, the request was sent to the second highest priority server group, in this way actually provides users with a way to heat back up.

2. Dynamic load balancing algorithm:

The minimum number of connections (Least Connection): transmitting the new server is connected to those for minimal connection processing.

The fastest response time (Fastest): transfer connection to the server in response to the fastest.

Observation (Observed): the number of connections and the response time of the server selection request for a new optimum balance of these two basis.

Prediction (Predictive): BIG-IP server using the current performance metrics collected, analysis was carried out, to select a server the next time slice, which will achieve the best performance of the server in response to user requests.

Dynamic allocation performance (DynamicRatio-APM): BIG-IP performance parameters collected applications and application servers, dynamically adjust traffic allocation.

3. Load balancing strategy Programmable control:

Load balancing via iRules control.

 

·END·

Siege lion of the road, the road is far away, the line will be to!

Guess you like

Origin www.cnblogs.com/w520/p/12082999.html