The subscriberd data represents the information about the serviced entities of a communication service provider, including the subscribers’ network identification and reference, associated payment model, balance and status with the applications to which they subscribed.
There has been a trend recently to perceive subscriber data management (SDM) as a big data problem. What makes it big data? Among other factors, it’s the complexity of the data stored, the velocity of data generation and the increasing volume of subscriber data. In addition, subscriber data needs to be shared concurrently with a number of software systems (the network, the application layer, and operational components), which all together raise a number of important challenges in the management of subscriber data:
- Long term evolution. Subscriber data is the backbone of any provider’s application grid. As its interface is implemented by a variety of network components, all of which are effectively in service, with a life span of 15-20 years, the SD repository needs to be prepared for the long run.
- In terms of model variability, as the data model is likely to grow and change, while relying on the existing data structure, the impact on an already scaled and measured system needs to be minimum and the attachment to existing data needs to be seamless.
- With regard to scalability, capacity planning must be an equally seamless process.
- High availability (redundancy, resilience with no visible performance drops, geographical distribution so the service runs regardless of a certain data center in trouble, automatic failover with no noticeable gap when a subsystem goes down or becomes available again)
- Real Time. The SD repository is on the service chain of real time services, who deal with requirements in the order of milliseconds
- Concurrency – the real time services described by the previous bullet are working concurrently.
- Interface exposure – providers often choose to expose the SD via the application layer to various 3rd parties who develop their own services
Several approaches to subscriber data management have been implemented, from the traditional silo approach to the directory structure, the solution used by Nokia Networks in developing One-NDS (Network Directory Server), one of the today’s top SDM products on the market.
Silo approach. In the traditional silo approach each new service requires a new application with its own database. That database needs to be put into a highly available architecture, integrated with a new interface into the provisioning systems and appropriate maintenance processes need to be set up. More importantly, the information across the silos needs to be synchronized, with a variable time sensitivity.
Database farm approach. In a “DB farm” approach, having a management layer over a variable number of fragments is a more modern solution, but there are still some concerns on how the overhead scales with the number of nodes and how the read speed is optimized.
Directory Structure based approach. A Directory is a repository of information, with a hierarchical tree-like structure, where data is stored as entries within the tree structure.
R Systems has successfully used and recommends the DS based approach for a number of features which can address the challenges of subscriber data management today. A Directory Service is …
… naturally optimized for read speed. DS is a hierarchical representation of data; accessing data is therefore equivalent to specifying the path (it doesn’t involve a lookup).
… easy to distribute / scale. A network DS relies on a self balancing mechanism, providing both redundancy and parallelism in data processing; adding nodes either to the front-end (the management layer) or to back end (the actual data repository) is a simple capacity planning task.
… able to accommodate eclectic data. The data is subject to more relaxed constraints than with the table-based DBMS.
… able to change schema on the fly. Changes to the existing data definition can be implemented without affecting the DS functions (as long as they do not affect the data integrity).
… able to patch additional functionality on top of existing system. Different schemas can link to the existing tree as they become part of the network logic.
… easy to integrate with a variety of applications, via LDAP or other standard protocols.
… backward compatible with the older GSM networks. In the absence of individualized SDM layers, HLR software can access the DS directly, since they were usually built on an LDAP accessible data structure.
… forward compatible with the upcoming LTE technologies, via a wide range of access APIs. Using a single SDM layer to serve multiple network domains has become one of the primary goals of network operators, with the bloom of new and varied technology options.
An illustration of this solution in use is one of the today’s top SDM products on the market: One-NDS (Network Directory Server) developed by Nokia Networks.
One-NDS is a high performance X.500 based directory, successfully deployed at the heart of multiple tier 1 network operators all over the world. It is highly distributed, has a hierarchical tree-like structure, a front end layer dispatching requests towards the back end nodes, and supports a variety of open interfaces, including LDAP, DAP, SQL, ODBC / JDBC.
For more details about the directory structure approach at the heart of SDM application, click here to see the presentation delivered by Valentin Raduti, System Architect – R Systems, at DevTalks 2015.