Page 440 - Handbook of Modern Telecommunications

Network Management and Administration 3-231

Source IP Address
NetFlow Key Fields Source Port
Destination IP Address
Destination Port
Layer 3 Protocol
TOS Byte (DSCP)
Input Interface

FIGu RE 3.9.4 Sensory data available within NetFlow records.

3.9.3.2 CDR-Based Sensor Networks
A substantial amount of valuable operational intelligence can be gathered by processing postsession
CDR (Call Detail Record) types of data. This is something of a special case for device-based sensor
networks, as the source of the data is usually the infrastructure nodes themselves. In IP networks, this
is often in service activity records such as IP Detail Record (IPDR) (which is also the basis of service
billing), flow records like NetFlow/IPFIX (from Cisco core routers or service platforms), JFlow (from
Juniper core routers), or statistically sampled equivalents such as sFlow. By their very nature, they
provide information about each subscriber session, including what application or service was invoked
and some basic statistical measures regarding the aggregate session, such as total traffic volume and
session duration and some basic quality metrics. Key information available from NetFlow records is
indicated in Figure 3.9.4.
CDR-based sensory deployments are very similar in nature to device-based deployments (see
Figure 3.9.4), in that all of the sensory data is drawn from network devices. The basic difference in this
case is that CDRs are generated by the devices and sent to a defined/configured destination; hence,
instead of using a polling collector to proactively gather the data, a receiver collector approach is used.
The primary advantage of using these types of data sources for service assurance is the fact that they
are being generated directly on the basis of each individual service invocation, and thus are true reflec-
tions of service session activities. Additionally, these are commonly available from IP device manu-
facturers at no additional cost (for the generation of the data records), although there is investment
required in tools for collecting and processing the records.
Disadvantages include the load that they place on the delivery infrastructure elements (though it
is small for most devices, even a small additional load may be deemed an unacceptable performance
risk), and the fact that this data set leaves out some of the most critical information that is required
to discretely recognize and understand individual subscriber sessions, complex multi-element content
services, and key quality metrics such as responsiveness and resolution within Web-based transport
protocols. Also, since CDRs are postsession records, they are not a practical basis for troubleshooting
live services in real time, or for recognizing intrasession anomalies.

3.9.3.3 Agent-Based Sensor Networks
The next category of technologies are those that measure service quality by using independent software
or hardware agents to generate and introduce synthetic test traffic into the service delivery network on a
periodic basis to assess service availability and responsiveness. This may be accomplished by deployment
of agentry directly by the operator, or by using a third-party service that executes the tests and reports
the results back to the operator. The test traffic can either be a select set of trials that exercise key service
components, representative approximations of complete service sessions, or replays of actual captured
subscriber sessions. There are also agent-based solutions that directly instrument live subscriber sessions
from the end device or at the edge of the network and passively monitor actual service sessions, though