If you are like most customers running time-sensitive applications such as SNA or real-time voice, you are probably obsessed with reducing delay in your frame relay network. Below is a list of areas where delay is introduced. (This issue focuses on real-time voice, as it is one of the most delay-sensitive applications.)
The first place delay may be introduced is in your CPE. As discussed in the last issue, the CPE must perform several tasks - such as prioritization, compression and fragmentation - each taking several milliseconds. In most cases, CPE delay ranges from 20 to 75 msec.
The next place delay is encountered is in the access line. This is often the slowest link the data will traverse, so it can be a major culprit in introducing network delay. The delay can be 5 to 50 msec, or even higher if your CPE must put voice in a queue for transmission across the access line. (Keep in mind that there are also access and CPE delays on the terminating side.)
As your traffic enters the frame relay network, each switch it must traverse will introduce delay. Many service providers are guaranteeing network delay of 100 msec. You may pay more for this low-delay service, but it's worth it in some cases. Overall, you should strive for a maximum delay of 250 msec, the point at which delay becomes noticeable to most people. Your users do not want their data performance to drop simply to achieve higher quality voice.
RELATED LINKS
Cisco buys QoS vendor
Network World, 5/5/98.
Demands for speed and quality of service drive evolution of frame relay services
Network World, 2/9/98.
The quality-of-service quandary
Network World, 4/13/98.
Sign up for the NWFusion Focus: Frame Relay newsletter
Additional NWFusion Focus: Frame Relay back issues
