The war between mice and elephants

被引:67
作者
Guo, L [1 ]
Matta, I [1 ]
机构
[1] Boston Univ, Dept Comp Sci, Boston, MA 02215 USA
来源
NETWORK PROTOCOLS | 2001年
关键词
D O I
10.1109/ICNP.2001.992898
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Recent measurement based studies reveal that most of the Internet connections are short in terms of the amount of traffic they, carry (mice), while a small fraction of the connections are carrying a large portion of the traffic (elephants). A careful study of the TCP protocol shows that without help from an Active Queue Management (AQM) policy, short connections tend to lose to long connections in their competition for bandwidth. This is because short connections do not gain detailed knowledge of the network suite, and therefore they are doomed to be less competitive due to the conservative nature of the TCP congestion control algorithm. Inspired by the Differentiated Services (Diffserv) architecture, we propose to give preferential treatment to short connections inside the bottleneck queue, so that short connections experience less packet drop rate than long connections. This is done by employing the RIO (RED with In and Out) queue management policy which uses different drop junctions for different classes of traffic. Our simulation results show that: (1) in a highly loaded network, preferential treatment is necessary to provide short TCP connections with better response time and fairness without hurting the performance of long TCP Connections; (2) the proposed scheme still delivers packets in FIFO manner at each link, thus it maintains statistical multiplexing gain and does not misorder packets; (3) choosing a smaller default initial timeout value for TCP Can help enhance the performance of short TCP flows, however not as effectively as our scheme and at the risk of congestion collapse; (4) in the worst case, our proposal works as well as a regular RED scheme, in terms of response time and goodput.
引用
收藏
页码:180 / 188
页数:9
相关论文
共 32 条
[1]  
ALLMAN M, 1998, 2414 RFC
[2]  
AMIR E, UCB LBNL VINT NETWOR
[3]  
BANSAL N, 2001, P ACM SIGMETRICS 200
[4]  
BERNET Y, 1999, FRAMEWORK DIFFERENTI
[5]  
BLACK S, 1998, 2475 RFC
[6]  
Bonald T., 2001, P ACM SIGMETRICS 200
[7]  
CARDWELL N, 2000, P IEEE INFOCOM 2000
[8]  
CHRISTIANSEN M, 2000, P ACM SIGCOMM 2000 S
[9]   Explicit allocation of best-effort packet delivery service [J].
Clark, DD ;
Fang, WJ .
IEEE-ACM TRANSACTIONS ON NETWORKING, 1998, 6 (04) :362-373
[10]  
CROVELLA ME, 1999, P 1999 USENIX S INT