Background load generation
With data services here and set to grow significantly during the coming years, how do operators ensure their upgraded data networks perform under load? We look at the issue of background load generation.
With the growing penetration of mobile data services, mobile operators must provide content and features that attract new customers whilst increasing ARPUs for existing subscribers.
Enhancing current infrastructure by integrating new functionality within their existing networks while maintaining QoS is a constant battle, but a task that operators must take seriously to grow and protect increasing data revenue streams, especially as the uptake of new services such as mobile TV and VoIP grows.
Impact of Upgrades
Many operators fail to implement background load generation during their testing activities; although up to four to five times more faults can be found just by adding this critical step during a testing process.
With an average of one major and five minor upgrades per vendor per year, the RoI on adding load test tools to an operator’s environment is quick to be realised, especially as undetected faults can cause serious problems that range from inaccurate billing information through to a complete network outage, lasting several hours. All of these result in lost revenues and increased churn from unhappy customers (particularly an issue for corporate customers with data SLAs).
Load testing has traditionally been performed by the Network Equipment Providers (NEPs) utilising their propriety tools which focus on stressing the Node to its published limits prior to general release. These tests are normally carried out in a controlled environment, limited to a standard set of “black-box” tests.
A Different Approach
To ensure stability in the live network operators require a different approach during their integration and verification, focused in three areas:
• Testing the Node configured with and operating within the operator network environment, connected to the operator’s subsystems.
• Using traffic models that represent real-life and realistic traffic profiles and customer usage patterns
• Performing pre-upgrade system integration and verification of new software releases and vendor swap-outs whilst simultaneously generating real-life background load.
This process also enables operators to test integration between their subsystems. Consider a scenario where an operator wishes to test 500 MMS/second towards their GGSN. Utilising a load test tool will not only test the GGSN node itself, but also the WAP gateway and MMSC.
In addition to delivering reliable data services, charging and billing is high on operators’ agendas. Load testing and background load generation enables network elements to be loaded and tested to ensure that CDR (Call Data Record) information is produced reliably and accurately. Data usage is increasing by the subscriber and so are its costs.
The assurance that they are being billed accurately is becoming a major focus of the new data subscriber.
Realistic Traffic Modelling
One of the major keys to success during testing lies in the continued development of valid and realistic traffic models. A model developed reflecting today’s data usage will be outdated in a matter of months. Traffic models should allow configuration of the key parameters such as Attached Subscribers and Rate, Payload sizes and Types, QoS and Inter/Intra system changes between 2G/3G networks. Each network element may require a different traffic profile strategy. SGSNs, for example, tend to require more testing under signalling load whereas GGSN testing normally focuses on payload and data services. Most operations and maintenance systems within operators should provide this information easily.
Ease of Use
One of the major contributing factors to an operator not traditionally performing such test is the lack product available. An operator needs a tool that will allow its own staff to quickly and effortlessly implement new traffic models focused towards their own network environment, without the complex scripting or protocol configuration commonly associated with tools produced by the vendor.
Easy-to-use tools and realistic traffic scenarios will not only allow background load generation using today’s traffic models but enable operators to “future proof” their networks by using virtual traffic models prior to releasing new features or promotions to the public (e.g. “How will my data network cope with the introduction of Video Streaming”).
Operators must embrace and implement background load generation in their processes from the initial test labs, to staging environments through to the first Node deployment. Only by doing this can they ensure they provide stable and reliable data services to their customers, reducing downtime and lost revenues.