Each site in a multisite deployment usually is interconnected by an IP WAN, or occasionally by a metropolitan-area network (MAN) such as Metro Ethernet. Bandwidth on WAN links is limited and relatively expensive. The goal is to use the available bandwidth as efficiently as possible. Unnecessary traffic should be removed from the IP WAN links through content filtering, firewalls, and access control lists (ACL). IP WAN acceleration methods for bandwidth optimization should be considered as well. Any period of congestion could result in service degradation unless QoS is deployed throughout the network. Voice streams are constant and predictable for Cisco audio packets. Typically, the G.729 codec is used across the WAN to best use bandwidth. As a comparison, the G.711 audio codec requires 64 kbps, whereas packetizing the G.711 voice sample in an IP/UDP/RTP header every 20 ms requires 16 kbps plus the Layer 2 header overhead. Voice is sampled every 20 ms, resulting in 50 packets per second (pps). The IP header is 20 bytes, whereas the UDP header is 8 bytes, and the RTP header is 12 bytes. The 40 bytes of header information must be converted to bits to figure out the packet rate of the overhead. Because a byte has 8 bits, 40 bytes * 8 bits in a byte = 320 bits. The 320 bits are sent 50 times per second based on the 20-ms rate (1 millisecond is 1/1000 of a second, and 20/1000 = .02).