Load-Balancing

BoostEdge Atlas offers extremely powerful layer-7 load balancing using a variety of methods. The load-balancing is used to support a bigger load by dispatching the requests across the servers. It is also used to ensure application's availability. In this latter case, the server health checking represents a key feature.

Servers are gathered in a cluster which is defined in a cluster profile. The requests are then dispatched across the servers belonging this profile and regarding their availability and their load.

Adaptive / Observed Dispatching Methods

When the load-balancing is actualy used to support a bigger load and relieve the application servers, the requests must be dispatched regarding the actual server load. Requests are then routed to the least loaded or iddle servers. Dispatching methods such as weighted round robin or weighted random exist in adaptive versions. The initial server weight is multiplied by an iddleness factor (iddleness = 1-load).

The server load is measured by either SNMP probes or is calculated regarding the server think time and the current pending requests.

Adaptive methods combine advantages of weighted methods and least loaded or fastest method preventing risks of pumping.

Availability Dedicated Dispatch Methods

In most of the case, the actual usage of the load-balancer is to ensure the application availability and business continuity. Therefore it makes sense for an load-balancer to have dispatch methods dedicated to this particular function. You then indicates a primary server, then a backup server then a secondary backup server and so on.

In such a case the architecture may not be symmetric. The primary server is a powerful one while secondary is a smaller configuration and the third one can be an ultimate backup virtual server.

Health Checking, a Key Feature

Since the load-balancing is mostly used to ensure the application availability, the health checking is a mandatory key feature. It is in charge of detecting whether or not a server is no longer working properly.
Read more ▶

BoostEdge's health checking goes far beyond "pingings" the server or even simply check the HTTP server respond to a request. Since the HTTP server may reply in a perfectly formatted (with a 200 OK status-code) that there is a critical data base failure, it is obvious that the content itself must be checked. Content that either must or must not be found in the the reply is indicated in a powerful and flexible way using regular expressions when necessary.

Session Tracking

Session tracking is required where information relating to the application session is held by servers and cannot be shared. Session monitoring can be based on hashing the source IP address, a private session cookie, an application cookie, the SSL session ID and server checking.

Clustering & Web Division

Also called web division or web partition. This involves bringing together functionally equivalent servers to create a cluster. Partitioning an application or a site is the first level of load distribution based on the URL. This type of distribution is called "URL based load-balancing". It is also the first stage in implementing load balancing. Once a cluster has been defined, you simply need to specify the method to be used to distribute the load in the cluster configuration.

Sorry Page

When the entire application plateform is down because either all servers are down or a common system such as the database server is down, BoostEdge can display a Sorry Page. This prevent the help desk or the network administrator to be overwhelmed by users complaints.