Sec.6.1 Introduction 497 Input rate 0.8 Limited buffer space Low capacity link C=1 Input rate High capacity A link C 10 B Low capacity link C=1 Low capacity link C=1 D (a) 1.8 Throughput when infinite buffer space is available indy6nojy at the central node 1.1 0.8 Retransmissions due to limited buffer space at the central node start here 0 1.0 Input Rate f of A to B Session (b) Figure 6.1 Example demonstrating throughput degradation due to retransmissions caused by buffer overllow.(a)For f approaching unity.the central node buffer is almost always full.thereby causing retransmissions.Because the A-to-B session uses a line 10 times faster than the C-to-D session.it has a 10-fold greater chance of capturing a free buffer and getting a packet accepted at the central node.As a result.the throughput of the A- to-B session approaches unity.while the throughput of the (-to-D session approaches 0.1.(b)Total throughput as a function of the input rate of the A-t0-B session. or number of hops traveled so far;at each node,separate buffer space is reserved for different classes,while some buffer space is shared by all classes. Proper buffer management can also help avoid deadlocks due to buffer overflow. Such a deadlock can occur when two or more nodes are unable to move packets due to unavailable space at all potential receivers.The simplest example of this is twoSec. 6.1 Introduction Input rate 0.8 497 Input rate f ::J C. .<: '"::J 1.1 :: .<: f- ro 0.8 0 f- A Limited buffer space '\, High capacity link C 10 Retransm issions due to limited buffer space at the central node start here B Low capaciw link C 1 (a) 1.8 ,---,,--' "". - ------- -- ...................... 1 .\ I \ Throughput when infinite buffer space is available at the central node I -~---r-~------------ I a 1.0 Input Rate f of A to B Session (b) Figure 6.1 Example demonstrating throughput degradation due to retransmissions caused by butler overflow. (a) For f approaching unity. the central node buffer is almost always full. thereby causing retransmissions. Because the A-to-B session uses a line 10 times faster than the C-to-D session. it has a 1O-fold greater chance of capturing a free buffer and getting a packet accepted at the central node. As a result. the throughput of the .4- to-B session approaches unity. while the throughput of the C-to-D session approaches 0.1. (b) Total throughput as a function of the input rate of the A-to-B session. or number of hops traveled so far; at each node, separate buffer space is reserved for different classes, while some buffer space is shared by all classes. Proper buffer management can also help avoid deadlocks due to buffer overflow. Such a deadlock can occur when two or more nodes are unable to move packets due to unavailable space at all potential receivers. The simplest example of this is two