Network Demarcation
Demarcation has to do with defining and protecting the borders between networks owned or managed by different entities while maintaining interconnectivity and interoperation of traffic and features between the two networks.
TDM PSTN gateways offered an implicit enterprise network demarcation point. Until recently VoIP has been deployed only in private enterprise and small business networks for on-net calls—calls that remain within the organization's own network. Off-net calls that went from the enterprise to (or from) the PSTN were converted between IP and TDM at the PSTN interconnect point (even though many service provider backbone networks have also been VoIP for many years).
With SIP trunking the provider-to-enterprise interconnect is now also migrating to using VoIP technology. This means you no longer need TDM PSTN gateways, but it also means you lose all the demarcation features TDM gateways implicitly provided to your network. These demarcation features include
- Compliance with service provider's User-to-Network Interface (UNI)
- Codec choice
- Fault isolation
- Statistics and voice quality reporting
- Billing and call accounting
- QoS marking
- Topology hiding (security)
These demarcation features are critical to the maintenance, security, and management of your network. An SBC, such as the CUBE, can be placed at the edge of your network to terminate the SIP trunk entry point and fulfill the needed demarcation role in an all-IP network connection. For smaller businesses CUCM Express might be deployed, which includes SIP trunk capability and border element demarcation features.
All the areas of demarcation are discussed in the remainder of this section, except topology hiding, which is further discussed in the "Security Consideration" section in this chapter.
Service Provider UNI Compliance
SIP trunk service providers offer an explicit UNI specification of what message types, formats, and fields are valid on their service offering. For the enterprise to comply with this UNI, it is often easier to place a border element at the edge of the network to normalize all the variants from different enterprise applications and endpoints than to try to configure each individual application or endpoint to comply with the UNI. It is especially true if the enterprise connects to multiple different SIP trunk providers, perhaps for least-cost routing or for redundancy purposes.
For the service provider, it is often easier to drop off a validated border element CPE device to ensure that the enterprise or small business network complies with its UNI—rather than certify each possible vendor and release combination of the possible applications and endpoints, the enterprise or small business might want to connect to their service.
Different deployment scenarios result from this need for network demarcation. Often enterprises want to manage their own border element so that they can control the configuration of this device and adjust it for new applications, application upgrades, or call flows. Alternatively, the service provider might provide a border element as a CPE device as part of the SIP trunk service to ensure UNI compliance regardless of the enterprise equipment. In some cases, especially for larger enterprises, both exist, and there is a pair of border elements at the enterprise edge, one side owned by the service provider (CPE), the other by the enterprise. This is separate from the SBC that always exists at the service provider edge and is a shared device among many SIP trunk customers.
The SIPConnect forum has been established by a consortium of members as an industry organization to focus on specifying and defining the SP UNI as a standard to ease some of interconnecting and interoperability issues currently still experienced.
Codec Choice
Codec choice was briefly discussed previously in this chapter in the "Transcoding" section. Transcoding is one of the demarcation features you might want to deploy in your network to normalize codec use at the border of your network to the choices you have engineered your network for, independent of the codec choices on the SIP trunk or those chosen by the offnet destination of the call. Newer wideband codecs, such as G.722, can be used in your network but might not yet be available on SIP trunks services. The choice of SIP delayed offer or early offer also has some influence over what codecs can be chosen for any particular call.
To control the use of codecs on your network to comply either with bandwidth engineering (call admission control) or with other enterprise policies, you have the following choices:
- Allow calls with inappropriate or incompatible codecs to fail.
- Involve a transcoder to resolve calls with incompatible codecs and to change incoming codecs to those you prefer to use on your network.
- Configure features to control codec negotiation and filtering in SIP call setups as the call passes through the border element.
Fault Isolation
In traditional TDM PSTN access, the PSTN gateway terminated the TDM connection from the provider's network and originated a VoIP connection inside your enterprise network. If voice quality or connectivity problems existed, this demarcation point was an easy place to conduct testing and isolate whether the problem existed within your enterprise network or whether it was the service provider's problem. TDM loop testing is common, enabling the service provider to test the TDM loop to the edge of your network to determine if the problem exists on that part of the connection.
Bringing a SIP trunk into the enterprise removes this demarcation point and, therefore, also the problem isolation techniques that existed for TDM interconnection. If voice-quality problems occur, it can be difficult to isolate whether they are caused by something in the service provider's network or by an element in your enterprise network.
Using a CUBE as an IP demarcation point restores this troubleshooting capability, enabling testing within the enterprise network up to the CUBE and testing from the service provider's side to the CUBE to determine where a fault might exist. The IP loop can be tested in the same conceptual manner (RTP loopback capability and Service Assurance Agent [SAA] responder support) as the TDM loop to allow the service provider to determine if the service is causing a problem or whether the problem exists in the enterprise.
Statistics
Metrics, such as delay, jitter, and voice quality scoring, help enterprises and providers monitor and control the voice quality on their networks. These metrics can typically be derived only at the endpoint (DSP) of a VoIP call and not in the middle of it. (No DSP is involved in the middle of the call.) For calls on SIP trunks, it is necessary to calculate, or estimate, some of these metrics at the border element to reflect the quality of the call on the enterprise side of the network, separate from the metrics of the call leg on the service provider side of the network.
You can use different features to derive a reading of the metrics at the network border. One is to use transcoding on the CUBE, which terminates the VoIP call leg on a DSP and re-originates it on the other side—because a DSP termination is involved, actual statistics on both call legs are available from the DSP.
Another method is snooping on the Real-Time Control Protocol (RTCP) statistics as they travel through the border and to report on some of these statistics. However, many VoIP endpoints do not support RTCP.
The IP Service Level Agreement (IP SLA) Real-Time Transport Protocol (RTP)-based VoIP operation feature provides another method to provide statistics. This Cisco IOS feature uses test calls to a DSP to determine values for voice quality metrics over different network segments. The CUBE can be either the originator or the destination of the IP SLA probes to provide readings for voice quality statistics up to your network border.
There is also the Cisco IOS Voice Performance Statistics on Cisco Gateways feature (using the command voice csr statistics) that collects call statistics such as active calls, failed calls, packet loss, latency, and jitter.
Billing
Typically, service providers bill without any information from the enterprise. Call detail records (CDR) from the CUBE can provide a consolidated aggregate view of calls sent and received on the SIP trunk and can be used to validate the service provider's billing.
Drawing billing records from your border element also provides a consolidated view of SIP trunk traffic use if you share a SIP trunk among multiple CUCM clusters or IP-PBXs.
Cisco IOS Software CDRs contain calling and called numbers, local and remote node names, data and time stamp, elapsed time, call failure class fields, and some vendor-specific attribute (VSA) fields. Each call through the CUBE is considered to have two call legs. Start and Stop records are generated for each call leg. These records can be sent to a RADIUS server or retrieved with Simple Network Management Protocol (SNMP) polling using the dial-control Management Information Base (MIB).
QoS Marking
TDM voice gateways originated IP packets and, therefore, could control the QoS markings on both the signaling and media VoIP packets entering your network for calls from the PSTN. With an end-to-end VoIP call over a SIP trunk, it's quite possible that the service provider preferred QoS markings are different from the ones you prefer, and, therefore, packets have to be remarked in both directions as the packets cross the border.
The CUBE is a back-to-back user agent and, therefore, has full control over packet marking in both directions and can be set either globally or based on destination. For example, if you have two SIP trunks to different providers and their choices of marking is different from each other and from your choice in the enterprise, the border element can remark these packets on a per-flow basis.
VoIP endpoints and call agents such as CUCM and CUCMExpress also have facilities to control and mark packets. These can be used directly if the enterprise markings are the same as the SP UNI markings, and an SBC can be used if markings need to be translated between the enterprise and the SP networks.