Improving Clustering Support in Radiator with Gossip Protocol

 Our Radiator is deployed in various operating environments, ranging from single server instances to large clusters, sometimes called server farms. Recently, we have developed Radiator support for Gossip protocol framework and Redis based Gossip implementation. The Gossip framework allows clustered Radiator instances to share information and event notifications. The instances can be standalone or part of a Radiator server farm.

One use case for the Gossip framework is co-ordinated detection of unreachable next hop proxies. Multiple Radiator instances can now signal next hop proxy unreachability and reachability information with Gossip messages. The instances may be part of server farm, completely separate processes running on the same or different hosts or any combination of thereof - depending on the customer’s network architecture.

Radiator 1 informs Radiator 2 via Gossip

This allows, for example, one server farm worker to run Status-Server queries when the FarmSize configuration parameter is set. When the member notices that the next hop has become unreachable, it can quickly alert the other members to immediately switch over to an alternate next hop. This provides a method to quickly recover from failures - and also switch back when the failure condition has been resolved.

In general, when the instances can share information, co-operating Radiator instances as a whole can function more efficiently and respond more quickly to configuration and other changes.

As our product development continues, we are planning to develop more clustering functionality to Radiator. Our aim is to provide even more performance to network environment with large amount of end users.

How can I use them as a Radiator customer

If you already use Radiator, see goodies/farmsize.cfg for a configuration example with shared duplicate cache and Gossip and Redis configuration. As our development work continues, more modules will be added and upgraded to use the Gossip framework.

If you are not yet a Radiator customer, please contact our sales team at sales(a)open.com.au with your question about. We are happy to help you with your issues.

Scaling and using Radiator in the cloud just became even easier

Infinite Lights, a photo by Jesus Solana (Creative Commons)

Radiator has long supported scaling with features such as server farm (FarmSize) and multiple load balancing algorithms. In the latest Radiator patches we introduce the first features in the series of new improvements aimed to make Radiator a perfect RADIUS/Diameter AAA component from private and telco clouds to public clouds such as Amazon.

As operators gradually move away from physical servers and appliances to architectures like telco clouds, network functions virtualisation (NFV) and software defined networking (SDN), the need for software components capable of scaling and communicating between each other increases. A modern AAA (authentication, authorisation, accounting) software component must have features for communication, organising cooperation (orchestration) and dynamic load balancing and that is where we started.

The current Radiator patches now include Gossip framework with an implementation built on top of Redis. Proxying and load balancing were enhanced to use Gossip for sharing information about next hop proxy reachability and unreachability. This provides fast detection and recovery from proxy failures for cooperating Radiator instances across several, even separate, servers. In addition to sharing information about network events, Gossip based duplicate cache is now available for server farm instances. This is, however, only the beginning. Moving towards Radiator 5.0, we are adding more features supporting new and future operator cloud functionalities. As always, patches will include these coming features first.

We fully recognise that we cannot do and figure out everything by ourselves and welcome your feedback. This helps us to continue developing Radiator to the component you want to use when building your services and products.

Interested in participating?

If you are an existing customer interested in these new features, please download the newest Radiator version and patches and try them. If you need help or want to give feedback, please use your existing support contract contact or send feedback via evaluation-support ( a ) open.com.au.

If you are interested in evaluating Radiator, please fill out our evaluation form and mention your interest in testing the cloud features. Our evaluation support team will contact you shortly.