Aptilo Networks has learned a great deal from all the 120+ customers that we have deployed carrier grade systems for over the past ten years. We have learned that future needs and applications may not be clearly foreseen by us today. What we can be sure about though is that they will come, and what we can presume is that they will surprise us. For this reason we have put a great emphasis on flexibility, scalability and high-availability in our new ALE architecture. Parametrization for the software and scalability for the hardware was the obvious focus in our system design.
Taking the scalability issue out of the equation
When it comes to performance we can only be sure of one thing. No matter if we use RISC-based, standard Intel-based or proprietary servers, the top performance of today will be considered a mediocre performance tomorrow. We can also be sure of that the price for server hardware will go down at the same rapid pace as the speed goes up.
The only way to be sure to take the scalability issue out of the equation is to obtain a linear scalability when adding more servers. With the fast development of server hardware it is important to go with a server platform that can offer the best price-performance-ratio. We firmly believe that the best scalability will be achieved by distributing our system over different layers running at different servers. We have chosen the industry standard Intel®-based hardware because we believe that it will provide the most cost-effective server platform for delivering additional performance by just adding another server. This architecture offers economy of scale thanks to its pure volume in number of units and the service provider will not risk doing large investments in individual high performance servers that may not fit the purpose at the end. True linear scalability can only be achieved with 100% flexibility in both hardware and software.
The new Aptilo ALE architecture
In the new Aptilo ALE architecture, the functionality is divided into three distinct layers:
- A management layer for configuration and monitoring
- A control layer for protocol-level interactions with external systems
- An execution layer for application logic processing.
The different layers are distributed over several server nodes to increase capacity where it is needed most and obtain high availability including geographic redundancy and distribution.
ALE application control layer
The control layer runs a number of protocol adapters that facilitate interoperability with the required set of external systems. These protocol adapters include RADIUS, Diameter, BGP (Border Gateway Protocol), SOAP/XML, LDAP and SNMP. New adapters are introduced as needed to implement new or modified relationships with external systems. This layer do not need to have redundant servers as it is state-less and can be replaced by any other server with the control layer running.
ALE application execution layer
All application logic is performed in this layer. It holds state and thus need to be on redundant server pairs. We have learned the hard way that no matter how much features we add there will always be one more needed. Every customer have different legacy aspects that we need to take into account when designing the solution. We have designed the ALE architecture based on the simple truth that there will always be new requirements and that neither our customers nor we have the time to wait for the next release. Thus new logic needs to be added as configuration, learn more about how we have solved this through the Aptilo ServiceGlue™ functionality. The application execution layer architecture allows us to run several different types of applications such as EAP-SIM authentication and Policy Management from the same scalable platform.
ALE application management layer
The configuration and monitoring is handled by the application management layer and it is tightly integrated with a management plane which spans over all layers. With this design we can assure that we add capacity where it is most needed. For signaling intensive applications it might be the control layer that needs most power while in applications with extensive business logic it might be the execution layer.
Scaling from one server system to large cluster systems
Aptilo’s products and solutions based on the Aptilo ALE architecture can for test purposes start with one redundant server pair running all required ALE layers and then scale all the way up to a large linearly scalable server cluster by just adding more server nodes.