Mediation–Take load off Billing

A service provider wanted to filter out non-billable calls. But his existing
mediation application utilized vendor-defined record normalization and
single-pass mediation methodology on input. According to the normalization
process, he didn’t have access to the right fields in the usage data to create
the filter definition. So he approached the vendor to get the desired filter
logic created, only to receive an estimate of six months for the development
process.

Ideally, the mediation application should have provided for creating the
filter logic in minutes time, through some graphical user interface–creation
of user-defined filter criteria should not be more than a series of mouse clicks
away.

The mediation application should not constrain the service provider in
reformatting an input record to a different output format. The record formatting
logic is driven through a selection of pre-defined functions. However, the
formatter allows multiple levels of complexity such as the ability to add
headers and trailers. Examples of formatting include AMA to EMI, conversion to
CIBER format, conversion of multiple input record formats to a single record
format required by a downstream application etc. Like Filter, Format has access
to all fields for maximum flexibility.

Validation is another key mediation issue. In addition to verifying field
content, the validation mediation feature can be used as a general auditing
tool. Users should be able to define the logic through the application GUI for
sequential receipt of usage data from any network element that provides some
sort of sequence number.

The mediation application should also provide for user-definable correlation
logic–to support wireless billing on distance as well as time by correlating
the cell sites of originating and terminating records.

Mediation in GPRS

Mediation of event data at a centralized point in the business process
provides the greatest efficiencies. Mediation enables each downstream
application to receive usage data in the format and manner it expects to receive
it. Changes in the network such as new network elements/services, additional
vendors, or record structure changes due to generic upgrades are all addressed
once, at the point of collection and mediation.

GPRS will require several enhancements to billing and mediation systems over
the existing ones. One of the core components that need to enhance is the
mediation device (MD). Some of the changes are directly related to GPRS, such as
the ability to handle different types of call detail records (CDRs). Others
involve system upgrades for better performance due to the larger volume of
incoming CDRs.

GPRS is an important step for the present-day GSM and TDMA network operators
who are in the migration path to 3G. It allows an operator to provide an
IP-based solution for data transfer to mobile subscribers. One of the plus
points of a GPRS network is that the radio resources are reserved only when they
are required and not for the entire session, as in circuit-switched voice
communications. A GPRS network can be implemented by overlaying on an existing
GSM/TDMA network two additional nodes: Serving GPRS Support Node (SGSN) and a
Gateway GPRS Support Node (GGSN).

Collecting CDRs

A mediation device acts as the initial point of contact for all OSS (IT)
systems that need to receive the unrated CDRs. It collects CDRs from all network
elements (switches, etc.) and acts as an intermediate storage buffer before
transferring the CDR data to downstream systems. In the case of a 2.5G network,
the device collects CDRs from MSCs and support nodes like SGSNs and GGSNs. It
collects charging information for each mobile station (MS) that it is serving.
It also collects usage of the radio interface for data transmission in mobile
originating and mobile terminating directions, as well as the usage or duration
of a packet data protocol (PDP) address. A PDP address is a network protocol
used by an external packet data network communicating with a GPRS network. An MS
can have a static or dynamic PDP address assigned by the network. This address
conforms to the standard addressing scheme–IPv6 or X.121–as used by the
network layer service.

GGSN is mainly responsible for collecting charging information related to the
service and usage of external packet data networks (identified by access point
name). A unique charging ID is generated for each activation of a new PDP
context (a set of parameters required for session management when an MS is
connected to an external packet data network). This information is held in the
mobile station, SGSN and GGSN for a PDP address. In order to receive or send
data, a PDP context needs to be activated.

As a subscriber moves between routing areas, the controls are handed over
from one SGSN to another. In such cases, the PDP context is transferred from the
first SGSN to the second one, and so are the charging IDs. Both SGSN and GGSN
collect charging information from same chargeable session/PDP context. A
composite key, comprising the charging ID and GGSN address, is needed to
assemble such records.

An optional charging protocol–GPRS Tunneling Protocol (GTP)–is designed
to facilitate CDR collection from SGSN and GGSN. GTP is used for tunneling
packet data in the GPRS network. The path protocol can be either UDP or TCP.

When the user moves to an area served by a new SGSN

At the end of a context

New Operational Needs

Mediation devices will be required to handle a variety of complex CDRs. In
addition, many new operational requirements will have a major impact on these
systems.

• Data collection at multiple nodes: In a pure voice
  GSM network, all CDRs are polled at an MSC–even though a network can have
  multiple MSCs (an operator can design to have all the CDRs polled at a
  single MSC). However, in case of a GPRS network, the CDRs need to be
  collected at more than just one node. In a GPRS network, based on routing
  area updates, context handling can be switched from an old SGSN to a new
  one. Thus charging records can be generated in more than one SGSN. In
  addition, data for a PDP context is also collected at the GGSN. Hence one of
  the main requirements of MD geared for a 2.5G network is the basic
  capability of collecting data from multiple nodes.

• CDR data feed for other systems: There has been a
  growing need for many systems to have an input of unrated CDRs. Fraud
  management systems would like to have the unrated CDRs for their analysis,
  as much as the rating and billing system. Based on business requirements,
  many operators would like a feed of unrated CDRs into the datawarehouse from
  the MD. Currently many operators use their rating and billing systems as the
  source of CDRs for other systems. However, in future, the MD would be
  treated as the sole source for unrated CDRs. The conclusion is that an MD
  having such a capability would have a niche in the market over its peers.

• Call aggregation: Currently in a pure voice
  network, aggregation or assembly of calls occurs mainly at the rating and
  billing system as one of the initial steps prior to rating. For instance, an
  MSC may generate two CDRs for a single call, if the call lasts for more than
  a certain amount of time. In a pure voice scenario, the number of cases
  requiring aggregation is minimal. However, with GPRS data calls, no CDR can
  be rated just by itself, so the load of aggregation would be considerably
  higher. And MD can reduce the load of the rating engine by performing
  initial aggregation.

• Duplication check: Most rating engines come with
  the capability for performing duplication checks so that a call is not rated
  twice. However, with the increase in volume of data, such checks would
  affect the performance. Though this is not an absolute necessity, it is a
  good feature for a MD to perform a first-level duplication check.

• Filtering CDRs: An operator may decide not to use
  all categories of CDRs for rating its subscribers. For instance, an operator
  may not use the M-CDRs for billing purposes. However, an operator may want
  to use them for fraud analysis. Since the MD would be treated as a source of
  IT multiple systems, it should be possible to have different filter criteria
  for filtering CDRs to different systems.

• Real-time mediation: As we strive for 3G, we also
  move closer to a real-time rating. In order to do so, it is but natural to
  have real-time mediation. An important factor makes real-time mediation even
  more necessary: access by customer service representatives (CSRs) to the
  CDRs in real time. For example, consider a user who has subscribed to a high
  level of QoS but who cannot really get a connection at the subscribed QoS.
  Dissatisfied with the network connection, he calls the CSR. The CSR will be
  able to look at the CDR generated for the session, and verify the QoS that
  the subscriber actually used. If it is inferior, the subscriber can get a
  discount. An operator will be able to provide such a service only if its
  mediation device worked in real time.

• Configuration file setup: Business requirements
  are pushing an operator to support real-time rating and mean that the MD
  must run 24x7. But if this is so, how can an operator make changes to the
  templates of the CDR files that it is mediating? The answer lies in having
  the MD contain a list of configuration files, with an operator managing
  different versions of it. An updated template could come into effect from a
  certain date and time. However, no operator will risk going ahead with a new
  version of a CDR template without testing it. This in turn requires the MD
  to be able to run in two parallel instances–one instance for processing
  the live CDRs, the other for testing with the configuration file.

• Roaming GPRS CDRs: Standards are evolving for GPRS
  roaming procedures: GRX or GPRS Roaming Exchange. One obvious question is:
  what are the implications of GPRS roaming for a billing system, and in turn
  for a mediation device. While roaming, the SGSN of the roaming network
  generates an S-CDR, and the HPLMN generates a G-CDR for the session that the
  subscriber initiates. While the G-CDRs are transferred to the home network
  right away, the S-CDRs are transferred from the visitor network to the home
  network, based on the normal roaming agreement–transfer account procedure
  (TAP). As an operator marches ahead, he will seek an MD capable enough to
  handle such CDRs.

Conclusion

The core difference between 2G and 2.5G/3G is their data handling capability.
The MD will require new features that incorporate GPRS CDRs and can handle voice
records. However, it is more than likely that an operator may receive additional
kinds of CDRs besides the ones noted in this article. For instance, if an
operator provides an intelligent network solution, as in prepaid, there would be
several other categories of CDRs.

Many features are essential for the mediation device to aggregate and
correlate GPRS CDRs correctly. That being the case, a specific business approach
for an operator would be to do weighted analysis of these requirements and
implement them in a convenient manner.

Medida Koteswara Rao Satyam
Computer Services

Call Data Records in a GPRS Network

Five kinds of call data records (CDRs) are collected by the MD from the GPRS
network.

S-CDR: When a mobile station (MS) performs a GPRS attach function with
an GPRS support node (SGSN), it sets up a link with the SGSN that is serving the
MS. GPRS attach is an operation by which an MS makes its presence known in the
network and which results in a location update procedure. The SGSN knows the
protocol data address used by the MS, and uses this address to route the data to
it.

SGSN generates an S-CDR for each activated PDP context. S-CDR collects
charging information related to packet data for a GPRS mobile. It stores
information such as served IMSI, served mobile subscriber integrated services
digital network (MSISDN), and GGSN address. It also stores the name of access
point of the external data network to which a GPRS MS is connected. Some of the
important fields in this CDR are the PDP type (X.25 or IP) and the PDP address
of the served IMSI (IPv4 or IPv6). Another important field is duration of a GPRS
call. The charging information is categorized by QoS.

The traffic data volume is categorized as uplink or downlink.
Each time QoS changes, the traffic data volume field is stamped. Some of the
conditions that can trigger the closure of an S-CDR are: termination of PDP
context, SGSN change, and upper limit reached on volume.

M-CDR: In order to better understands the generation
of an M-CDR, it is important to know how the concept of mobility management (MM)
applies in a GPRS scenario. Mobility management refers to the possible states of
a subscriber (more specifically, a mobile station).

The three possible states are idle, standby and ready. In an
idle state a GPRS MS is not reachable, since the MS and the SGSN don’t contain
any location information for the GPRS MS. So data transmission to and from the
GPRS MS is not possible.

In a standby mode, a GPRS MS can receive signaling
information from the network. However, it will still not be able to transmit any
data.

But when the MS is ready to receive data, and acknowledges
it, the state automatically changes to ready, and it is now ready to both
receive and send data. Now the MM context stores location information up to the
cell level. After the transmission is over, the MS can still be in a ready state
as long as the timer for the ready state (as set by an operator) doesn’t
expire. When it does, the MS moves back to a standby state. A timer governs the
standby state as well. Once this timer expires, a network can initiate a GPRS
detach procedure, and the MS would shift to the idle state.

SGSN plays a core role in mobility management. It also
collects information about mobility management of a GPRS mobile. When a GPRS
mobile performs a GPRS attach function at the SGSN, an M-CDR is generated at the
SGSN. This CDR contains information such as served IMSI, served MSISDN, and the
SGSN address. A change in the routing area would stamp the change of location
field. Since mobility management doesn’t deal with the actual data
transmission, there won’t be any field related to volume of data traffic in an
M-CDR. Factors that may lead to the closure of an M-CDR are: SGSN change, GPRS
detach, and upper limit reached on duration.

G-CDR: GGSN is the first point of contact for an
external packet data network that interconnects with a GSM network supporting
GPRS. It contains routing information for each attached GPRS MS. Once a GGSN
receives an incoming data request, it uses the location information to route
data transfer to the SGSN serving the MS.

GGSN generates a G-CDR to collect charging information
related to packet data. G-CDRs are very similar to S-CDRs. However, they will
not store any information about the specific cell location of the MS, such as
the local area code. On having an SGSN change, a G-CDR needn’t be closed, but
the address of the new SGSN is added to the list of SGSN addresses.

S-SMO-CDR and S-SMT-CDR: In a traditional 2G scenario,
an SMS message is transmitted over the control channels, thus limiting the
length of the message to 160 characters (for GSM). When a short message is
transmitted over the GPRS radio channels, it doesn’t have those restrictions.
An interesting case arises if an MS is both IMSI-attached and GPRS-attached. In
that case it is up to the operator to configure whether the message should be
sent through GPRS radio channels or the non-GPRS control channels.

When an MS sends a short message over GPRS channels, the SGSN
is the initial contact point to receive the message and pass it on to the short
message center. In case of a mobile terminating message, once the SGSN serving
the MS gets a notification from the SMS-GMSC (SMS-Gateway MSC) of an incoming
short message, it passes the message to the MS over the GPRS radio channels. The
SGSN generates an S-SMO-CDR for a mobile-originated short message, and an S-SMT-CDR
for a mobile-terminated SMS message.