Cisco’s three steps for assuring QoS for voice

Recently we talked with Cisco about its three-step approach to implementing appropriate quality of service for voice.

We spoke with Craig Cotton, manager of product marketing, and Graham Gudgin, enterprise manager technical engineer at Cisco.

The first step in Cisco’s approach is to determine what is required on the LAN and WAN to assure QoS for the voice portion of the shared network. Once the correct levels of voice-over-IP bandwidth are decided, the engineers set aside bandwidth that can be statistically reserved for call admission.

Subscription rates are enforced using the call admission control in the gatekeeper. Cisco’s call manager enforces QoS for H.323, MGCP, SCCP (a.k.a. Cisco Skinny) and CTI/TAPI/JTAPI.

Note that Cisco doesn’t provide “overflow” to the public switched telephone network (PSTN) on a call-by-call basis unless the over-subscription rate is reached. Once the network is up and running, Cisco does support reroutes to the PSTN when an “oversubscribed” caller tries to place a call.

Step two is to determine what is required to support QoS for data applications, using profiles and setting aside the bandwidth required. Cisco has data on packet size and latency requirements for a variety of data applications, including SAP, Oracle, e-mail, Lightweight Directory Access Protocol-enabled databases, PeopleSoft, and other common applications. Application requirements will vary from site to site and are an integral part of proper planning.

Step three is to measure and monitor both the data and voice applications. Information is pulled from router queues, providing information on a per-class or per-flow basis. This post-planning step allows corrective action as data and voice requirements change.

Next time, we’ll summarize our series and provide some great reference sites on QoS planning.