Avaya 555-245-600 사용자 설명서
Fragmentation
Issue 6 January 2008
327
Thus, the complete configuration for Frame Relay traffic shaping looks like:
map-class frame-relay NoBurst
no frame-relay adaptive shaping
frame-relay cir 384000! (for a 384K CIR)
frame-relay mincir 384000
frame-relay be 0
frame-relay bc 3840
interface serial 0
frame-relay class NoBurst
Fragmentation
One large cause of delay and jitter across WAN links is serialization delay, or the time that it
takes to put a packet on a wire. For example, a 1500-byte FTP packet takes approximately
214 ms to be fed onto a 56-Kbps circuit. For optimal voice performance, the maximum
serialization delay should be close to 10 ms. Thus, it can be problematic for a voice packet to
wait for a large data packet over a slow circuit. The solution to this problem is to fragment the
large data packet into smaller pieces for propagation. If a smaller voice packet comes in, it can
be squeezed between the data packet fragments and be transmitted within a short period of
time.
The sections that follow discuss some of the more common fragmentation techniques.
takes to put a packet on a wire. For example, a 1500-byte FTP packet takes approximately
214 ms to be fed onto a 56-Kbps circuit. For optimal voice performance, the maximum
serialization delay should be close to 10 ms. Thus, it can be problematic for a voice packet to
wait for a large data packet over a slow circuit. The solution to this problem is to fragment the
large data packet into smaller pieces for propagation. If a smaller voice packet comes in, it can
be squeezed between the data packet fragments and be transmitted within a short period of
time.
The sections that follow discuss some of the more common fragmentation techniques.
MTU
The maximum transmission unit (MTU) is the longest packet (in bytes) that can be transmitted
by an interface without fragmentation. Reducing the MTU on an interface forces a router to
fragment the large packet at the IP level. This allows smaller voice packets to squeeze through
in a timelier manner.
The drawback to this method is that it increases overhead and processor occupancy. For every
fragment, a new IP header must be generated, which adds 20 bytes of data. If the MTU is
1,500 bytes, the overhead is approximately 1.3%. If the MTU is shortened to 200 bytes,
however, the overhead increases to 10%. In addition, shortening the MTU to force
fragmentation increases processor utilization on both the router and the end host that needs to
reassemble the packet.
For these reasons, shortening the MTU is only recommended as a last resort. The techniques
described later in this section are more efficient, and should be used before changing the values
of the MTU. When changing the MTU, size it such that the serialization delay is less than or
equal to 10 ms. Thus, for a 384-kbps circuit, the MTU should be sized as follows: 384000 bps
by an interface without fragmentation. Reducing the MTU on an interface forces a router to
fragment the large packet at the IP level. This allows smaller voice packets to squeeze through
in a timelier manner.
The drawback to this method is that it increases overhead and processor occupancy. For every
fragment, a new IP header must be generated, which adds 20 bytes of data. If the MTU is
1,500 bytes, the overhead is approximately 1.3%. If the MTU is shortened to 200 bytes,
however, the overhead increases to 10%. In addition, shortening the MTU to force
fragmentation increases processor utilization on both the router and the end host that needs to
reassemble the packet.
For these reasons, shortening the MTU is only recommended as a last resort. The techniques
described later in this section are more efficient, and should be used before changing the values
of the MTU. When changing the MTU, size it such that the serialization delay is less than or
equal to 10 ms. Thus, for a 384-kbps circuit, the MTU should be sized as follows: 384000 bps