 |
This layer of the
system is perhaps the most key ingredient in making this system work.
It allows each web server to be redundant by adding servers into a server
farm, or a group of servers, assigned to do a specific task. This means
if one web server goes down for any reason or stops serving up requests,
then the others within the server farm will take over its responsibility.
It is like sharing the load of the web site on many servers. With the
addition of load balancing, the administrator can scale the web site's
infrastructure to meet the demand of the site.
The load balancers
maintain distribution to the server farm in many different ways. The goal
is constant, to distribute the load equally among all servers in the server
farm and maintain 24/7 uptime. Each load balancer accomplishes this task
by using proprietary algorithms. In many cases the most valuable algorithm
is the least connected user algorithm. In this configuration, the destination
of each request (or server in the server farm) is polled and the one that
has the least currently connected user receives the next request. For
example, if 5 servers each have 10 users connected to them it would mean
that each server had 2 users currently connected. As one server completes
the request and sends the user's data back onto the Internet, the connection
is dropped, leaving one server with only 1 user attached. In this example,
the next request would then go to that server fulfilling the algorithm
and maintaining equal distribution among all the web servers.
|
 |