Network Magic

Network management can take everyday network-generated data and transform it into the answers to your questions.

People responsible for managing telecom networks dream of the day when all of the information that their networks generate can be thrown into a hat and then, like a magician, they can pull out whatever they need, whenever they need it — a list of radios that need tuning, the customers most likely to churn next week or a list of call destinations that are being misrouted. Much of the data that is thrown into the hat, gets swirled around and then pulled out in a different form belongs under the cloak of network management. This is a term that is defined differently depending upon whom you are talking to and the network's stage of sophistication.

Reporting Data

Network management feeds on the raw data generated by countless intelligent devices that are part of wireless networks or monitoring devices that are attached to network elements or to network connections.

Reporting of information is not fully standardized, so network-management companies have to provide probes that can interpret the stream of data from each device and turn it into an intermediate form prior to digestion by their network-analysis routines.

A considerable amount of filtering, reduction and combining of data is necessary because of the huge volume produced, but also because small amounts of data from different pieces of equipment may, together, tell much more than when they are examined individually. Only a system that understands the topology of the network can tell which pieces of data should be combined, which should be kept as individual items and which should be summarized.

Why Bother?

It pays to ask why network-management data should be collected at all. According to John Kairupan, TTI Telecom (www.ttitelecom.com) sales manager, network management has become an integral part of profitability. This is because a properly managed network will not only keep its customers happy, but it also will be able to balance its resource inputs. Knowing when to send a crew out to a cell site to prevent a major outage can save a major account, and when there is one crew available and two cell sites in trouble, it is important to know which one is going to have the biggest bottom-line impact if it goes out of service.

Moving Up the Chain

Network management usually is visualized as a pyramid. (See Figure 1 on page 40.) At the bottom is data originating from “managed elements,” devices that are intelligent enough to report data when abnormal events happen (alarms) or, for monitoring purposes, on a regular basis. This data is fed to the element-management layer, which can interpret the data and thereby reduce the number of interfaces that upper layers have to deal with.

The element-management layer still is dealing with fairly low-level events. It is up to the network layer to map them onto the network topology and assimilate them into more meaningful terms. Filtering and consolidation of data is important. If an alarm is reported 100 times, it may be meaningful, but humans certainly do not want to see all of those occurrences streaming by on their screens. The network layer may be able to examine them, conclude that a microwave link is noisy and change its status on a screen, perhaps generating a single alarm once a threshold is passed.

Higher levels in the chain of network management are more recent additions and will become even more important as beleagured telecom carriers attempt to retain the faith of stock markets by squeezing every drop of profitability out of their networks. The business layer attempts to correlate network data in terms that are important to the business, such as revenue being generated by the network and losses created by outages. Looking at the network in this way encourages prioritization of problem solving based on the financial impact to the company.

Feedback

There is no point in gathering and condensing network data if nothing is done with it. At some point, the system will detect an anomaly, and a piece of equipment will need to be taken out of service for maintenance or replacement. There is some disagreement about whether this process should be manual or automatic. Raghu Shankar, ADC Metrica (www.metrica.com) marketing manager, is adamant that customers want to have information presented to them in an actionable form, but they want a human to press the button to initiate a reaction. Kairupan disagreed, saying that automatic reactions are necessary, but that often a human would be asked to give the “go/no go” approval to the action suggested by the network-management system.

Both viewpoints are probably right. Systems are too complex for humans to fully analyze. Semi-formulated decisions, at least, must be presented to humans. On the other hand, no software is yet smart enough to completely take over running the network.

Within the network-management domain, information about new problems is continually bubbling to the top. Decisions are made by humans and their machine assistants, and information flows back down to help mitigate or rectify the problems.

Data From Call-Detail Records

Call-detail records (CDR) represent a huge source of additional data, a source that is much more customer oriented than information produced by managed elements. Some companies specialize in analysis of this data that can be used for more than billing and fraud management.

Barry Murphy, Openet Telecom CEO (www.openettelecom.com), emphasized that it's important to intercept and analyze CDR prior to their use for billing because much of the information needed for network management is superfluous for that purpose. By doing this, not only can companies like his provide a consistent interface to many CDR formats, but they also can filter the data for network-management purposes, as well as for other purposes such as churn management. Information about the problems being experienced in calls can refine the interpretation of problems reported by the network hardware.

Link Monitoring

Data links do not generate information about themselves; that is something that only the devices connected to the links can do. Well, that is not quite true. Some companies, notably Agilent (www.agilent.com) and Coherent Networks (www.coherent.com), specialize in passive monitoring equipment that watches every bit flow across the network, decoding the protocols and monitoring for packet errors and retransmission. In the case of SS7 links, monitoring can occur at the top of the protocol stack. Daniel Doherty, Coherent vice president of sales, noted that this allows diagnosis of problems based on point-to-point traffic analysis. Situations can occur when every call from a wireless carrier to a certain destination is failing because of a provisioning problem in an intermediate network. Without the ability to monitor at this level, the problem would have to be discovered by analyzing customer complaints.

Service-Level Agreements

It is increasingly common for major clients of wireless networks to have service-level agreements (SLAs), which may be embedded in an agreement to maintain a virtual private network.

Smart carriers do not monitor SLAs as an output of the billing process because then problems are detected only in time to pay discounts to customers. IN-management systems should monitor SLAs continually, alarming not when the agreements are violated, but when they are in danger of being violated. Even if there is a violation, it is better to be pro-active and let customers know when the problems occur, rather than letting individual complaints flood customer service.

SLA monitoring can use information from CDRs to identify groups of subscribers that are being hit hardest by problems, as well as the network topology. A cell site serving a business park may, for example, have much more effect on SLAs when outages occur than a cell site in a more residential area.

RF Optimization

RF optimization would seem to be outside the scope of network management. It is, after all, something done by specialized engineers with colored maps on paper or on high-resolution screens. Channels are grouped by frequency and apportioned in the way that maximizes capacity and minimizes bad things such as interference. However, according to Israeli company Schema (www.schema.com), this is all wrong, and optimization is a continuous process that is an integral part of network management. Radio channels are simply network resources, and every last drop of useful information should be extracted from the mound of data that is generated on a regular basis or through routine monitoring activities.

A lot of data that comes out of element management reflects radio problems. Other information comes from the traditional technique of drive testing, although this is not always reliable because drive testing usually is done when the network (and roads) are quiet, but the system should be optimized for use during the busiest times of the day. Schema also can get information from handsets themselves. Digital handsets have the ability to monitor the signal strength of multiple cell sites in a process known as mobile-assisted handoff. Being able to use this information is like having thousands of drive testers who are most active during the busiest times and, best of all, do not know that they are testing, so they do not have to be paid. Schema would not describe how it knows where the mobile is, which is obviously necessary to identify places of poor coverage.

Schema takes all of this data and, through analysis, not only can identify problems in the network today but also can predict where problems will occur in the future and identify where to place new channels.

Many companies view network management as a product: a set of software and equipment that allows carriers to monitor their networks. However, for some carriers, network management is a service that they are happy to contract out. Fredrik Winterlind, Ericsson vice president of business management & marketing (www.ericsson.com), described how Ericsson first works with the customer to define rules and policies and then builds what is described as a service-assurance system to match. Under circumstances such as these, even a big equipment vendor such as Ericsson has to be able to work with many other vendors, even those that are normally fierce competitors, to be able to provide a seamless system to its customers. This mode of operation, according to Winterlind, is particularly common for carriers that are new in the business or those that anticipate being acquired before being in business for too long.

Looking to the Future

According to Lisa Ludwig, Ericsson director of product marketing, 3G systems raise new challenges for network management not only because of the new technologies involved, but also because of the integration (or collision) of two world views — telephony and data communications, neither of which understands the other.

Network management allows the efficient use of the capital investment in a network. Gone are the days when the stock market was willing to buy into any plans for new networks. Company executives are now much more focused on generating money from within or reducing losses; network management can make a big contribution to both. If Congressman Anthony Weiner (D-NY) has his way, the U.S. government will start regulating service standards for wireless consumers, further increasing the need for network management.

---

Crowe (crowed@cnp-wireless.com) is a wireless-standards consultant and editor of Cellular Networking Perspectives, a wireless-standards and -technology bulletin.