[February 21, 2002]

Pros And Cons Of Data Collection Techniques In Service Providers' Complex Environments

BY JONJIE SENA

Network data mediation has traditionally been the "black box" of a service provider's business -- a background process that is typically handled by a mediation vendor. In today's complex and highly competitive market, however, service providers are taking greater interest in this aspect of their business, and for good reason. Industry journals and conferences contain more and more reports of mediation "rescues" in which inefficient systems are updated or replaced by robust and intelligent solutions that recover thousands of dollars per day in billing revenue that had previously slipped through the cracks.

Revenue leaks and performance losses can occur in any network that has difficulty integrating information from newer technologies such as Internet Protocol (IP) services, or where the mediation system is inflexible and cannot easily accommodate multiple formats and protocols. As service providers acquire, or form partnerships with, other service providers, the mediation and OSS systems can become stressed and inefficient, potentially resulting in a loss of revenue and service quality.

For this reason, when seeking to enhance or upgrade their systems, service providers should not only understand their own networks, but the choices proposed by mediation vendors. Perhaps the best place to begin understanding a mediation vendor's approach is right on the front line of the mediation process -- data collection.

Data Collection
In establishing a data collection strategy for downstream operations and support systems, the mediation vendor has two primary data collection methods to choose from: probes and direct interfaces (DI) to network elements (NE). Probes are passive data-gathering hardware devices that "spy" on the network, recording all data that passes through the point in the transmission medium where the probe resides. The DI method, on the other hand, uses a data-exchange mechanism to retrieve selected data directly from the network element.

Before examining the pros and cons of each method, it is important to understand the kinds of data that such methods collect, as the data type is the most important factor in determining the best collection method to use.

Consider the example of watching a movie via online streaming video -- in this case there are two streams of data involved. The first is the actual movie content (i.e., the sound and video) which is called the media stream. The second stream contains information about the movie being watched, such as the name or reference number of the move, who is watching it (e.g., the subscriber number or IP address), the length of the movie, the director's name, and so on. This is the usage stream, also known as the accounting stream.

The media stream is always available while a service is being used, because the media stream is the actual service being delivered (whether it is a streaming movie, a telephone call, or a file download). The usage stream, on the other hand, is not always available, depending on the type or configuration of the network element involved. Also, the media stream can be quite large, but the usage stream is typically significantly smaller, requiring less bandwidth, storage space, and processing.

Pros And Cons Of Probe Technology
The main advantage to using probes is that they provide a network element-independent interface to capture data. For example, data sent across a TCP/IP network uses standard headers that allow a probe to quickly identify, at a very high level, the kind of traffic and payload, without requiring an interface with the network element. As such, probes can be deployed quickly and easily in environments that use standard protocols.

Another advantage with probes is that all the data transmitted for a particular service is available to the mediation and OSS systems. In other words, the OSS can process the entire media stream, which in some cases may be necessary to obtain a complete picture of the service provided. For example, there may be no accounting data generated at all, in which case the usage information can only be derived by processing and summarizing the media stream.

On the other hand, the main disadvantage of using probes is that such a large amount of data is captured -- data that needs to be stored and processed. This may be overkill; an OSS does not need to hear the entire telephone conversation to be able to account for it or to provide statistics regarding the usage.

Another disadvantage to using probes is that wire-speed data capture is difficult, especially as the bandwidth of the network increases. If the same services are deployed on faster networks, then hardware upgrades for the probe will probably be required. Also, the probe needs to listen to every link in the chain to get the full picture, instead of just one network element. The typical approach is to attach probes to the end points in the chain, which usually -- but not always -- gives sufficient information for most purposes.

Pros And Cons Of DIs To Network Elements
Capturing all data generated by (or passing through) a network element requires a lot of bandwidth and processing, and in many cases provides far more data than is necessary. Unless one wants to know the language used, or how many times a particular word is used, it is generally sufficient to base OSS functions on the accounting information -- calling party, called party, time of day, call type, resources used, quality of service, facilities utilized, etc. -- alone.

Capturing information directly from a network element provides accounting data that is intimately connected to the system that generates it. It captures information about the idiosyncrasies of the service and implementation provided by the network element -- no other data source provides more usage information. It's like hearing a story first-hand instead of piecing it together from multiple secondary sources. As such, DI technology provides the most complete accounting information available in the network.

Direct interfaces are usually software driven, at least for data and next-generation network elements. However, this is straightforward and uncomplicated for mediation vendors who have already created other interfaces for whatever type of switch is involved. If the same services are offered on faster networks, no changes are needed for mediating the usage information via direct interfaces.

On the other hand, each network element potentially has its own proprietary interface, which may require custom development. This may result in a linear increase of interfaces to be developed as new network elements are introduced. In practice, however, this is rarely a problem. Most vendors provide the same interface for most of their network elements, so once an interface for a particular switch model is available, all models from the same vendor are automatically supported. Many also adopt standard protocols, such as FTP, FTAM, BX.25, SNMP, Q3, etc. Once an interface is built, it can usually be reused for other network elements.

Some people argue that the DI method is risky because the technology intermittently interrupts switch functions and utilizes processing capacity when responding to a request for usage data. However, these interruptions are usually low-priority events with no discernable affect on switch function or network traffic, and a properly implemented system will not cause dropped calls or other switch problems.

Probes And DI In Complex Network Environments
In the search for the ideal mediation solution, it must be recognized that no one solution addresses all mediation requirements. Rather, the mediation vendor should utilize the method that is most appropriate for the case at hand. Industry analysts such as the Yankee Group agree:

The Yankee Group feels that a successful NDM [network data mediation] platform will encompass many different collection technologies, such as embedded software and probes. The gathering of information in a data environment is extremely complex, encompassing pieces of information from several disparate network elements. The general industry consensus is that no single data source or protocol is optimal for encompassing the OSS/BSS information needs. Some tout network probes as a possible solution for tapping the information stream; however, the Yankee Group feels -- as do many vendors -- that probes are only one component of an overall NDM collection strategy and are appropriate only in certain situations. (Yankee Group Report, Vol. 2, No. 6, June 2001)

In complex network environments, the best choice is often a mixed strategy. A service provider seeking a mediation solution should keep these factors in mind, and in particular they should look into the experience and expertise of the mediation vendor. The best solution is to be found where the vendor not only understands the technology they are selling, but has a thorough understanding of the client's network environment and makes decisions based on each unique situation.

Jonjie Sena is a Convergent Mediation product manager for ACE*COMM Corporation. ACE*COMM is a global leader in advanced Convergent Mediation products and enterprise telemanagement applications for wired and wireless voice, data, and Internet communications providers.