Cisco DSP Resources for Transcoding Conferencing and MTP
This chapter provides information about how Cisco digital signal
processor (DSP) resources are used for transcoding and conferencing. The
modules, which are available for use with
Cisco Unified Communications Manager, can perform conferencing, Media Termination
Point (MTP), and transcoding services in addition to serving as a PSTN gateway.
DSP resources on the Cisco gateway, for example, Catalyst 4000 (WS-X4604-GWY), Catalyst 6000 (WS-6608-T1 or WS-6608-E1), Cisco 2600, Cisco 2600XM, Cisco 2800, Cisco 3600, Cisco 3700, Cisco 3800, or Cisco VG200, provide hardware support for IP telephony features that are offered by Cisco Unified Communications Manager. These features include hardware-enabled voice conferencing, hardware-based MTP support for supplementary services, and transcoding services.
Verify with your Cisco account manager the devices that support conferencing, media termination points, and transcoding services.
The DSP resource management (DSPRM) maintains the state for each DSP channel and the DSP. DSPRM maintains a resource table for each DSP. The following responsibilities belong to DSPRM:
Discover the on-board DSP SIMM modules and, based on the user configuration, determine the type of application image that a DSP uses.
Reset DSPs, bring up DSPs, and download application images to DSP.
Maintain the DSP initialization states and the resource states and manage the DSP resources (allocation, deallocation, and error handling of all DSP channels for transcoding and conferencing).
Interface with the backplane Protocol Control Information (PCI) driver for sending and receiving DSP control messages.
Handle failure cases, such as DSP crashes and session terminations.
Provide a keepalive mechanism between the DSPs and the primary and backup Cisco Unified Communications Managers. The primary Cisco Unified Communications Manager can use this keepalive to determine when DSPs are no longer available.
Perform periodic DSP resource checks.
When a request is received from the signaling layers for a session, the system assigns the first available DSP from the respective pool (transcoding or conferencing), as determined by media resource groups and media resource group lists, along with the first available channel. DSPRM maintains a set of MAX limits (such as maximum conference sessions per DSP or maximum transcoding session per DSP) for each DSP.
A switchover occurs when a higher order Cisco Unified Communications Manager becomes inactive or when the communication link between the DSPs and the higher order Cisco Unified Communications Manager disconnects. A switchback occurs when the higher order Cisco Unified Communications Manager becomes active again and DSPs can switch back to the higher order Cisco Unified Communications Manager. During a switchover and switchback, the gateway preserves active calls. When the call ends, the gateway detects RTP inactivity, DSP resources release, and updates occur on the Cisco Unified Communications Manager.
Introducing the WAN into an IP telephony implementation
forces the issue of voice compression. After a WAN-enabled network is
implemented, voice compression between sites represents the recommended design
choice to save WAN bandwidth. This choice presents the question of how WAN
users use the conferencing services or IP-enabled applications, which support
only G.711 voice connections. Using hardware-based Media Termination Point
(MTP)/transcoding services to convert the compressed voice streams into G.711
provides the solution.
The MTP service can act either like the original software MTP
resource or as a transcoding MTP resource. An MTP service can provide
supplementary services such as hold, transfer, and conferencing when the
service is using gateways and clients that do not support the H.323v2 feature
of EmptyCapabilitiesSet. The MTP, provided by the Cisco IP Voice Media
Streaming Application service, can be activated as co-resident with
Cisco Unified Communications Manager or activated separately without
Cisco Unified Communications Manager. Both of these services operate on the
Cisco Unified Communications Manager appliance (server). The Cisco IP Voice Media
Streaming Application service installs as a component with
Cisco Unified Communications Manager; however, for a dedicated MTP server, the
Cisco CallManager service would not be activated (only the Cisco Voice IP Voice
Media Streaming Application service).
When MTP is running in software on
Cisco Unified Communications Manager, the resource supports 48 MTP sessions. When
MTP is running on a separate
Cisco Unified Communications Manager appliance (server), the resource supports up
to 128 MTP sessions. In addition, Cisco Voice Gateway Routers also can provide
Observe the following design capabilities and requirements
for MTP transcoding:
Provision MTP transcoding resources appropriately for the number
of IP WAN callers to G.711 endpoints.
Each transcoder has its own jitter buffer of 20-40 ms.
The following summary gives caveats that apply to MTP
Make sure that each
Cisco Unified Communications Manager has its own configured MTP transcoding
If all n MTP transcoding sessions are utilized, and an n + 1
connection is attempted, the next call will complete without using the MTP
transcoding resource. If this call attempted to use the software MTP function
to provide supplementary services, the call would connect, but any attempt to
use supplementary services would fail and could result in call disconnection.
If the call attempted to use the transcoding features, the call would connect
directly, but no audio would be received. If a transcoder is required but not
available, the call would not connect.
You can configure voice compression between IP phones through the use of regions and locations in Cisco Unified Communications Manager. However, the Cisco Catalyst conferencing services and some applications currently support only G.711, or uncompressed, connections. For these situations, MTP transcoding or packet-to-packet gateway functionality provides modules for the Cisco Catalyst 4000 and Cisco Catalyst 6000. A packet-to-packet gateway designates a device with DSPs that has the job of transcoding between voice streams by using different compression algorithms. For example, a user on an IP phone at a remote location calls a user at the central location. Cisco Unified Communications Manager instructs the remote IP phone to use compressed voice, or G.729a, only for the WAN call. If the called party at the central site is unavailable, the call may roll to an application that supports G.711 only. In this case, a packet-to-packet gateway transcodes the G.729a voice stream to G.711 to leave a message with the voice-messaging server.
Voice Compression IP-to-IP Packet Transcoding and Conferencing
Connecting sites across an IP WAN for conference calls
presents a complex scenario. In this scenario, the modules must perform the
conferencing service as well as the IP-to-IP transcoding service to uncompress
the WAN IP voice connection. In the figure below, a remote user joins a
conference call at the central location. This three-participant conference call
uses seven DSP channels on the Catalyst 4000 module and three DSP channels on
the Cisco Catalyst 6000. The following list gives the channel usage:
Cisco Catalyst 4000
One DSP channel to convert the IP WAN G.729a voice call into
Three conferencing DSP channels to convert the G.711 streams
into TDM for the summing DSP
Three channels from the summing DSP to mix the three callers
Cisco Catalyst 6000
Three conferencing DSP channels. On the Cisco Catalyst 6000,
all voice streams get sent to single logical conferencing port where all
transcoding and summing takes place.
Figure 1. Multisite WAN Using Centralized MTP Transcoding and Conferencing
IP-to-IP Packet Transcoding Across Intercluster Trunks
Intercluster trunks connect Cisco Unified Communications Manager clusters. Intercluster trunks allocate a transcoder on a dynamic basis.
The Cisco Catalyst 6000 module uses the MTP service regardless of whether transcoding is needed for a particular intercluster call. Cisco Unified Communications Manager supports compressed voice call connection through the MTP service if a hardware MTP is used.
The following list gives intercluster MTP/transcoding details:
Outbound intercluster calls will use an MTP/transcoding resource from the Cisco Unified Communications Manager from which the call originates.
Inbound intercluster call will use the MTP/resource from the Cisco Unified Communications Manager that terminates the inbound intercluster trunk.
For compressed callers, you can accurately provision the MTP transcoding resources.
Hardware-Based Conferencing Services
Hardware-enabled conferencing designates the ability to
support voice conferences by using DSPs to perform the mixing of voice streams
to create multiparty conference sessions. The voice streams connect to
conferences through packet or time-division-multiplexing (TDM) interfaces.
The network modules, depending on the type, support both
uncompressed and compressed VOIP conference calls. The modules use Skinny
Client Control Protocol to communicate with
Cisco Unified Communications Manager to provide conferencing services. When the
conferencing service registers with
Cisco Unified Communications Manager, it announces that only G.711 calls can
connect to the conference. If any compressed calls request to join a
Cisco Unified Communications Manager connects them to a transcoding port first to
convert the compressed call to G.711.
Observe the following recommendations when you are
configuring conferencing services:
When you are provisioning an enterprise with conference ports,
first determine how many callers will attempt to join the conference calls from
Cisco Unified Communications Manager region. After you know the number of
compressed callers, you can accurately provision the MTP transcoding resources.
Conference bridges can register with more than one
Cisco Unified Communications Manager at a time, and
Cisco Unified Communications Managers can share DSP resources through the Media
Resource Manager (MRM).
The PSTN gateway and voice services module for the Cisco
Catalyst 4003 and 4006 switches supports three analog voice interface cards
(VICs) with two ports each or one T1/E1 card with two ports and two analog
VICs. Provisioning choices for the VIC interfaces include any combination of
Foreign Exchange Office (FXO), Foreign Exchange Station (FXS), or Ear &
Mouth (E&M). Additionally, when configured as an IP telephony gateway from
the command-line interface (CLI), this module can support conferencing and
You can configure the Cisco Catalyst 4000 voice gateway
module in either toll bypass mode or gateway mode; however, you can configure
the module conferencing and transcoding resources only in gateway mode. Gateway
mode designates the default configuration. From the CLI, you can change the
conferencing-to-transcoding ratios. After the gateway mode is enabled, the 24
DSPs (4 SIMMs with 6 DSPs each) for the module occur as described in the
Over the PSTN gateway using G.711 only-96 calls
In a G.711 conference only-24 conference participants; maximum of
4 conferences of 6 participants each
Unlike the WS-X6608-x1, which can mix all conference call
participants, the Cisco Catalyst 4000 WS-X4604-GWY module sums only the three
dominant speakers. The WS-X4604-GWY dynamically adjusts for the dominant
speakers and determines dominance primarily by voice volume, not including any
The Cisco Catalyst 4000 conferencing services support G.711
connections only, unless an MTP transcoding service is used.
Transcoding to G.711-16 MTP transcoding sessions
The following information applies to the Cisco Catalyst 4000
The WS-X4604-GWY uses a Cisco IOS interface for initial device
configuration. All additional configuration for voice features takes place in
Cisco Unified Communications Manager.
The WS-X4604-GWY can operate as a PSTN gateway (toll bypass mode)
as well as a hardware-based transcoder or conference bridge (gateway mode). To
configure this module as a DSP farm (gateway mode), enter one or both of the
following CLI commands:
The WS-X4604-GWY requires its own local IP address in addition to
the IP address for
Cisco Unified Communications Manager. Specify a loopback IP address for the local
Signaling Connection Control Part.
You can define a primary, secondary, and tertiary
Cisco Unified Communications Manager for both the conferencing and MTP transcoding
Cisco Catalyst 6000 WS-6608-T1 or WS-6608-E1
The WS-6608-T1 (or WS-6608-E1 for European countries)
designates the same module that provides T1 or E1 PSTN gateway support for the
Cisco Catalyst 6000. This module comprises eight channel-associated-signaling
(CAS) or primary rate interface (PRI) interfaces, each of which has its own CPU
and DSPs. After the card is added from
Cisco Unified Communications Manager as a voice gateway, you configure it as a
conferencing or MTP transcoding resource. Each port acts independently of the
other ports on the module. Specifically, you can configure each port only as a
PSTN gateway interface, a conferencing node, or an MTP transcoding node. In
most configurations, configure a transcoding resource for each conferencing
Whether acting as a PSTN gateway, a conferencing resource,
or an MTP transcoding resource, each port on the module requires its own IP
address. Configure the port to have either a static IP address or an IP address
that the DHCP provides. If a static IP is entered, you must also add a TFTP
server address because the ports actually get all configuration information
from the downloaded TFTP configuration file.
The following figure shows one possible configuration of the
Cisco Catalyst 6000 voice gateway module. This diagram shows two of the eight
ports of the module as configured in PSTN gateway mode, three ports in
conferencing mode, and three ports in MTP transcoding mode.
Transcoding for G.711 to G.729a/G.729ab/GSMFR-128 sessions
Transcoding for G.711 to G.729/G.729b/GSM EFR-96 sessions
For a software MTP (DSP-less with same packetization period for both
devices supporting G.711 to G.711 or G.729 to G.729 codecs), 500 sessions can
occur per gateway; for a hardware MTP (with DSP, using G.711 codec only), 200
sessions can occur per NM-HDV2 and 48 per NM-HD.