Cluster LoadBalancer Main Page | Features | Cluster LoadBalancer FAQ

Turbolinux Cluster LoadBalancer 10 is an enterprise-class solution that allows you to leverage your existing network resources to create scalable and reliable services. With it, you can significantly improve quality-of-service levels for virtually every TCP/IP network service, including web, email, news, and FTP. Cluster Server provides the architectural framework that will allow your network to effortlessly grow to meet new demands.

Cluster LoadBalancer implements load balancing and fail-over support of network services. Load balancing allows the services to run on multiple systems. The cluster will distribute client connections among the servers that make up the cluster. Fail-over allows the service to run on a single server. If that server should fail, another server within the cluster will take over for it.

You can think of Cluster LoadBalancer as similar to RAID. Whereas RAID uses an array of disks, Cluster Server uses an array of servers. Both provide the same features: enhanced speed, reliability, redundancy, and scalability. Cluster Server distributes the workload among several servers instead of concentrating all the work on one large server. However, the cluster will appear as a single machine to clients accessing it.

There are several hardware solutions available that perform the same function as Turbolinux Cluster LoadBalancer. These closed boxes tend to be very expensive and less flexible. By using a Linux-based system, you have finer control of the cluster. You also have the option of running other services on the cluster manager, and can have the cluster manager double as a cluster node. Cluster Server also allows redundancy of the traffic manager itself, so you do not have a single point of failure like many of the hardware-based solutions.

Cluster Server is a high-performance solution. The traffic management takes place at a very low level within the kernel. While all incoming traffic must come through the traffic manager, outbound traffic can go from the cluster node directly out to the client. Because most TCP/IP services have larger replies than requests, this is an important optimization.

In addition to forwarding traffic, Cluster LoadBalancer monitors the health and availability of the network resources. It continuously samples all server nodes, verifying that the applications are running properly. This is accomplished through the use of intuitive application polling agents. In addition, each backup traffic manager repeatedly queries the master traffic manager in order to verify that the cluster itself is functional.

Target Audience
Turbolinux Cluster LoadBalancer 10 is targeted at medium to large companies who want to implement the high availability or scalability features at a modest price. Internet Service Providers will find the product useful to provide a higher level of uptime as well as scalability that allows them to add servers to the cluster to improve performance. Large enterprises can use the product to deliver standards-compliant services to large numbers of clients, either internally or on the Internet. Medium-sized companies can use the software to leverage existing computer systems as the company's needs grow.

An administrator implementing Cluster Server should be familiar with Linux or UNIX and have a good understanding of TCP/IP networking. While clustering is a fairly simple concept, the implementation details can be rather complex. Troubleshooting any problems that arise will require not only understanding the concepts behind TCP/IP, but also experience with the real-world problems that can arise.

Turbolinux Cluster LoadBalancer 10 is not a Beowulf cluster, and is not intended to compete with Beowulf. It is not used to cluster CPU-bound processes, but instead focuses on network-based services.