Student: Shih-Chiang Tsao Advisor: Ying-Dar Lin Date: 2007/12/12 Dissertation Dissertation Fairness Controls for TCP-equivalence at Endpoint and Request-Response Scheduling at Gateway Where are bottlenecks? What kind of fairness? How and where to control it?
52
Embed
Student:Shih-Chiang Tsao Advisor:Ying-Dar Lin Date:2007/12/12
Dissertation. Fairness Controls for TCP-equivalenceat Endpoint and Request-Response Schedulingat Gateway. Student:Shih-Chiang Tsao Advisor:Ying-Dar Lin Date:2007/12/12. Where are bottlenecks? What kind of fairness? How and where to control it?. Bottlenecks for the Internet Traffic. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
1/Aggr(m)= the time taken by a scheme to increase its rate with a factor of m.
timeLast loss
ratem
1
???
2 3 4 5m
200
400
600
800
1000
1/Aggressiveness(RTTs)
WARC(160) SIMD(1/16)
GAIMD(
1/5,1/
8)
TCP
IIAD(1,2/3)
Fast as the most aggressive scheme
[Definition] [JGM03]
24 /482007/12/12
Better Tradeoff between Smoothness and Aggressiveness
0.02 0.04 0.06 0.08
500
1000
1500
2000
2500
3000
3500
SIMD
Smoothness (CV[w])
1/Aggressiveness (RTTs)
GAIMD
IIAD
WARC
WARC(160)SIMD(1/16)GAIMD(1/5,1/8)IIAD(1,2/3)
0.02 0.04 0.06 0.08
500
1000
1500
2000
2500
3000
3500
SIMD
Smoothness (CV[w])
1/Aggressiveness (RTTs)
GAIMD
IIAD
WARC
WARC(160)SIMD(1/16)GAIMD(1/5,1/8)IIAD(1,2/3)
445
150
1080
25 /482007/12/12
Analysis on Responsiveness[Definition] [JGM03]1/Resp(m) = the number of loss events required by a scheme
to decrease the rate with a factor of m.
.
Smoothness (CV[w])
0.02 0.04 0.06 0.08 0.1 0.12 0.14
10
20
30
40
50
0.02 0.04 0.06 0.08 0.1 0.12 0.14
10
20
30
40
501/Responsiveness
(#loss events)
WARC(160)SIMD(1/16)GAIMD(1/5,1/8)IIAD(1,2/3)
WARC
1/Responsiveness
(#loss events)
2 4 6 8m
10
20
30
40
50
WARC(160)K=3,N=12
GAIMD(1/5,1/8)IIAD(1,2/3)
SIMD(1/1
6)
TCP=GAIMD(1,1/2)
1/Responsiveness
(#loss events)
2 4 6 8m
10
20
30
40
50
WARC(160)K=3,N=12
GAIMD(1/5,1/8)IIAD(1,2/3)
SIMD(1/1
6)
TCP=GAIMD(1,1/2)
2 4 6 8m
10
20
30
40
50
WARC(160)K=3,N=12
GAIMD(1/5,1/8)IIAD(1,2/3)
SIMD(1/1
6)
TCP=GAIMD(1,1/2)
more losses for smoothness
26 /482007/12/12
Probability of False-Positive Enabling HR
AssumeX(-j) is an i.i.d. exponential distribution
forms a gamma distribution (n, λ) 1( )
n
jX j
Invoked when the mean of inter-loss time does not change
P=10-3 ->False Positive per 1000 losses35 mins when W=5~30, RTT=50~300ms [JID04]
2.2 2.4 2.6 2.8 3 3.2 3.4k
0.002
0.004
0.006
0.008
0.01
N=16 N=12
N=10N=8
Prob.
K
10-3
2.2 2.4 2.6 2.8 3 3.2 3.4k
0.002
0.004
0.006
0.008
0.01
N=16 N=12
N=10N=8
Prob.
K
10-3
0.3 30
0.05 6
1 11000 2100
0.3 0.05 30 6 1.5
WRTT dW dRTT
1
1 1
( , ) ( ,1)
( ) ( , )
( ) [ ] ( )
( ) ( ),
TCP K
N N
j j
gamma N gamma N
P R N R t s
N NP X j EX P X j
K KN N
F FK K
27 /482007/12/12
Fairness Test for TCP-Equivalence: Under the Variant-Losses Network
WARC:Average fixed #
of CWND
WARC:Average fixed #
of CWND
28 /482007/12/12
Fairness Test for TCP Equal-Share15 Mbps-link
60 Mbps-link
TimeoutMechanism
TimeoutMechanism
Equal share
Equal share
WARC
WARC
TEAR
GAIMDSIMD
29 /482007/12/12
Fast Aggressiveness & Responsiveness
1.0
1.2
1.4
1.6
1.8
2.0
2.2
TFRC TEAR WARC GAIMD SIMD WARCw/o HR
norm
. #
of
loss
es
WARC decreases rate with fewer losses
WARC decreases rate with fewer losses
Fast aggressiveness:WARC and SIMD
TFRC
TEAR
WARC w/oOne-RTT reduction
GAIMD TCP
20sec
30 /482007/12/12
Smoothness over Different Time Scale
WARC is smooth as TFRC
WARC is smooth as TFRC
Smoother rate than
TCP
Smoother rate than
TCP
SIMD
GAIMD
SQRT
IIAD
Better smoothness
Better smoothness
TEAR
(0.1 sec)
31 /482007/12/12
Low Start-up Latency for Constrained Streaming(e.g. video conference)
late packets
WARC has low ratio of late packets
WARC
TCP
TCP
WARC
32 /482007/12/12
Applicability of TCP-equivalent Smooth Rate Control
s
IP
UDP
Socket
APP
RTP/RTCP
IP
UDP
Socket
APP
RTP/RTCP Rate Control
User-layer Solution
(IETF Draft)
LiveMedia Library (LGPL), DirectShow RTP Filter
IP
TCP
Socket
APP
Rate Control
Kernel-layer Solution
(RFC4340, S. Floyd)
IP
DCCP
Socket
APP Layered/Base Protocols
Supported in Linux Kernel
A possible solution in MS Windows
Datagram CongestionControl Protocol (DCCP)
33 /482007/12/12
Summary WARC
RTE control model + Fixed number of CWNDs Fairness, Aggressivness,
History-reset mechanism Responsiveness
TCP-equivalence and TCP equal-share Fairness under stationary loss condition.
For non-periodic loss conditions Fast Aggr. & Rspo. for drastic change
Smoothness
Problems on Applying Fair Queuing Discipline to Schedule Requests at Access Gateway
for Downlink Differential QoS
No-monthly fee solution for downlink differential service
35 /482007/12/12
Where to Schedule Packets?
InternetGUGI
W1
W2GG
W3
User-side gateway
H1
Hn
ISP-side gateway
access link
Uplink requests ->
<- Downlink responses
User-side gateway (GU) or ISP-side gateway (GI) ? GU is bought by the user’s specification and easy to be managed GI is owned by ISP. Additional charge may requrie. Packets are not queued at GU GI cannot see the IPs of H1~Hn
Scheduling uplink requests at GU
to managing downlink responses Class-based Fair Queuing
Queuing packets
36 /482007/12/12
monopolizes the link bandwidth
sending one-by-one
Responses share the downlink neither is appropriate
sending reqs one-by-one sending a request right after g
etting a response
1. Time to Release the Next ...Packet Request
monopolize
packets
simultaneous
responsesS
Srequests
37 /482007/12/12
Selecting in the order of service-completion time Known packet size
Fairness should rely on response size
Response size is unknown until it returns
2. From Which Queue to Release the Next ..
Packet Request
SS
requests
8 7 4
9 5 3
6 2 1Q1
Q2
Q3
packets
1i
i i LF F
known packet length
? ? ?
? ? ?
? ? ?Q1
Q2
Q3
requests
response
response size is onlyavailable in 1st
packet of response
38 /482007/12/12
3. User-based Weighted Fairness Class-based
Between different types of traffic: e.g. voice or ftp
Admission Control User-level Differentiation
High-class users get more bandwidth than low-class users
39 /482007/12/12
Cr
Q2
Qn
Cq
Requests
Response
requestselector
SC1
SCn
W
End of RI
U
Wmax
Minimum-Service First Request Scheduling (MSF-RS)
End of Rsp.
Minimum-service order arbiter (MOA)Q1
A changes BA BA is referenced by BA B
A A is a variable
Data flow
Window-basedrate controller
(WRC)
requestreceiver
UC1 UCn
requestreleaser
w1 w3
Minimum-Service First Request Scheduling
SC: Service CounterUC: User Counterw: Weight
InternetGUGI
H1
Hn
40 /482007/12/12
Minimum-service Order Arbiter (MOA)
Cr
Q2
Qn
Cq
Requests
Response
requestselector
SC1
SCn
W
End of RI
U
Wmax
End of Rsp.
Minimum-service order arbiter (MOA)
Q1
Window-basedrate controller (WRC)
requestreceiver
UC1 UCn
requestreleaser
w1 wn
2. Select from the class with the min SC
2. Select from the class with the min SC
1. Log the amount of received service
1. Log the amount of received service
1
kk k ii i
i i
LSC SC
w UC1
kk k ii i
i i
LSC SC
w UCthe length of the
received response k in bytes
41 /482007/12/12
Q2
Qn
Cqrequestselector
SC1
SCn
Minimum-service order arbiter (MOA)
Q1
requestreceiver
UC1 UCn
w1 w3
Window-based Rate Controller (WRC)
Requests
ResponseCr
W
End of RI
U
W+
End of Rsp.
Window-basedrate controller (WRC)
requestreleaser
Release requests if W<W+
Release requests if W<W+
W : the number of outstanding requestsT : the time interval between two updatesSi : the responses in bytes received during TC : the link capacity. K: a constant. Ui: the link utilization. U+: upper bound of U
1 min{ , }i ii
UW K W
U
1 min{ , }i i
i
UW K W
U
/ T,
Ci
i
SU
/ T,
Ci
i
SU
42 /482007/12/12
Analysis of User-perceived LatencyLong queuing time of request+ Short transmission time of response
= Short user-perceived latency
Client send
request
Gateway
getrequest
Gateway
sendrequest
Gatewayget
response
Clientget
response
User-perceived latency
Tq Ts
Ta=Tq+Ts
Time
0 1 2
(4*1+4*2)/8=1.5
1111
2222
Example20 40 60 80 100
m
0.5
0.6
0.7
0.8
0.9
W+=10W+=20
W+=40
W+=80
TaMSF-RS/Ta
ordinary
43 /482007/12/12
S
Qi
Qj
sublink1
sublinkW+
Rspi,1
Rspi,W+
+ji
i j
LW L
w w
Time
Normalized Service
Class i
Class j
t0 t1 t2
0
i
i
W L
w
Analysis for Worst-case Fairness of MSF-RS:
+ji
i j
LW L
w w
1 21 2( , )( , ) ji
i j
D t tD t t
w w
Di(t1,t2): the responses receved by Class i in bytes between t1 and t2
wi: the weight of Class iW+ : # of sub-linksL+ : Resp. of max. size
Fairness Parameter defined by Golestani for analysis of SCFQ
44 /482007/12/12
Weighted Fairness and Sharing
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1 2 3 4 Phase
BW(kbps)
Class 1
Class 2
Class 3
All
0
50
100
150
200
250
300
4 8 12 16 20 24
Hosts in Class 1
Ban
dwid
th (
Kbp
s)
Class 1
Class 2
Class 3
0
50
100
150
200
250
300
4 8 12 16 20 24
Hosts in Class 1
Ban
dwid
th (
Kbp
s)
Class 1
Class 2
Class 3
0
400
800
1200
1600
2000
0 100 200 300 400 500 600 700 800
Time (sec)
BW (kbps)
BW of class 1
BW of class 2
BW of class 2
Bandwidth Sharing
Bandwidth Sharing
Weighted FairnessWeighted Fairness
Ban
dwid
th p
er u
ser
(Kbp
s)
Class-basedUser-based
Users in Class 1 Users in Class 1
BW1<BW2 or BW3
45 /482007/12/12
2.95
6.78
11.91
1.34
1.68
1.50
6.768.83
0
4
8
12
16
20
Sec
transmission time
Q time
Class 1 Class 2 Class 3 Average No Scheduling
User-Perceived Latency Lower congestion Lower transmission time
Client send
request
Gateway
getrequest
Gateway
sendrequest
Gateway
getrespons
e
Clientget
response
Higher # of concurrent connectionsHigher loss rate
46 /482007/12/12
Experimental Results User-perceived Latency
Squid Squid with MSF-RS
ms/request 1686.1 1174.9
(includes queuing time 515.5)
Throughput
2 Mbps 10 Mbps
10 Classes MSF-RS Squid 22.4 28.17
100 Classes MSF-RS Squid 23.01 29.02
Original Squid 31.61 42.45
Lower CPU loading due to fewer concurrent transactions in MSF-RS
Percentage on CPU utilization
user space127.0.0.1:3128
kernel spaceLinux
MSF-RSSquid
Realistic Servers in Internet
eth0eth1
192.168.2.53
Port Redirect (iptables): iptables -t nat -A PREROUTING -i eth1 -s 192.168.2.0/24 -p tcp --dport 80 -j REDIRECT --to-port 3128
Rate Limiting by switch: Input 2Mbps / Output 2Mbps
switch
Avalanche(Clients)
user space127.0.0.1:3128
kernel spaceLinux
MSF-RSSquid
Realistic Servers in Internet
eth0eth1
192.168.2.53
Port Redirect (iptables): iptables -t nat -A PREROUTING -i eth1 -s 192.168.2.0/24 -p tcp --dport 80 -j REDIRECT --to-port 3128
Rate Limiting by switch: Input 2Mbps / Output 2Mbps
switch
Avalanche(Clients)
47 /482007/12/12
Summary for MSF-RS Scheduling uplink requests -> Control Downlink Responses MSF-RS= Minimum-service Order Arbiter (MOA) +
Window-based Rate Control (WRC)
User-based weighted fairness Bandwidth Sharing among classes Reduce 20~30% of user-perceived Latency Reduce 25% of CPU loading
Low congestion Fewer concurrent transactions
48 /482007/12/12
Dissertation ConclusionsPublic Fairness:1. Taxonomy and evaluation of 8 TCP-friendly schemes
TCP-equivalence and TCP equal-share Rate-based fairness +
historical/super-linear aggressiveness + fixed history responsiveness TFRC: if meeting TCP-compatibility is the major concern SIMD: if fast aggressiveness is favorable
2. The design of WARC RTE control -> Non-periodic Fairness, Fast aggressiveness as SIMD History-reset procedure -> Fast Responsiveness as TFRC Better Meeting TCP-equivalence and TCP equal-share Smoothness in short-term for interactive constrained streaming
Private Fairness: The design of MSF-RS
Scheduling Uplink Requests to Manage Downlink Responses User-based Weighted Fairness High utilization while reducing 30% of user-perceived latency Reducing 25% of CPU loading
49 /482007/12/12
References[FF99] Floyd, S., and Fall, K., “Promoting the Use of End-to-End Congestion Control in the Internet.” IEEE/ACM Transaction
s on Networking, August 1999.[PFT98] J. Padhye, V. Firoiu, D. Towsley, and J. Kurose, “Modeling TCP throughput: A simple model and its Empirical Validat
ion,” in Proc. of ACM SIGCOMM’98, Sep. 1998, pp. 303-314.[YL00] Y. Yang and S. Lam, “General AIMD Congestion Control,” in Proc. of IEEE ICNP 2000, Nov 2000, pp. 187-198.[FHP00] S. Floyd, M. Handley, J. Padhye, and J. Widmer, “Equation-based Congestion Control for Unicast Applications,” in
Proc. of ACM SIGCOMM’00, Aug 2000, pp. 43-56.[ROY00] I. Rhee, V. Ozdemir, and Y. Yi, “TEAR: TCP Emulation at Receivers-Flow Control for Multimedia Streaming,” Tech.
Rep., Department of Computer Science, NCSU, Apr. 2000.[JGM03] Shudong Jin, Liang Guo, Ibrahim Matta, and Azer Bestavros, "A Spectrum of TCP-friendly Window-based Congestion
Control Algorithms," IEEE/ACM Transactions on Networking, vol. 11, no. 3, pp. 341-355, June 2003.[BB01] D. Bansal and H. Balakrishnan., “Binomial Congestion Control Algorithms,” In Proc. of IEEE INFOCOM, April 2001,
pp. 631-640.[PKT99] J. Padhye, J. Kurose, D. Towsley, and R. Koodli., “A Model Based TCP-friendly Rate Control Protocol,” In Proc. of N
OSSDAV, June 1999.[BCC98] B. Braden, D. Clark and J. Crowcroft, “Recommendations on Queue Management and Congestion Avoidance in the Int
ernet,” RFC 2309, Apr 1998, Available at http://www.ietf.org. [BBF01] D. Bansal, H. Balakrishnan, S. Floyd, S. Shenker, “Dynamic Behavior of Slowly-responsive Congestion Control Algorithms,” in Proc. of ACM SIGCOMM’01, Aug. 2001, pp. 263-274.
[LT03] Lahanas A and Tsaoussidis V, “Exploiting the efficiency and fairness potential of AIMD-based congestion avoidance and control,” Computer Networks, vol 43, no 2, pp. 227-245, Oct. 7 2003.
[VB05] Milan Vojnovic and J.-Y. Le Boudec, "On the Long-Run Behavior of Equation-Based Rate Control," IEEE/ACM Transactions on Networking, vol. 13, no 3, pp. 568-581, June 2005.
[ZDP01] Y. Zhang, N. Duffield, V. Paxson, and S. Shenker, “On the Constancy of Internet Path Properties, “ in Proc. of ACM SIGCOMM Internet Measurement Workshop, Nov. 2001, pp 197-211.
[CLB04] Gwyn Chatranon, Miguel Labrador, and Sujata Banerjee, "Fairness of AQM Schemes for TCP-friendly Traffic" in Proc. of IEEE Globecom, Dallas, Dec. 2004, pp 721-731.
50 /482007/12/12
[AAB05] E. Altman, K. Avrachenkov, and C. Barakat, “A Stochastic Model of TCP/IP with Stationary Random Losses,” IEEE/ACM Transactions on Networking, vol. 13, no 2, pp. 356-369, April 2005.
[NS06] The Network Simulator - ns-2, 2.1b9, http://www.isi.edu/nsnam/ns/.[RHE99] R. Rejaie, M. Handley, and D. Estrin, “Rap: An End-to-end Rate-based Congestion Control Mechanism for Real-time Str
eams in the Internet,” In Proc. of IEEE INFOCOM, Mar. 1999, pp. 1337-1345.[SW00] D. Sisalem and A. Wolisz, “Lda+ tcp-friendly adaptation: A measurement and comparison study,” in Proc. of NOSSDAV,
June 2000.[WDM01] J. Widmer, R. Denda and M. Mauve, “A survey on TCP-friendly congestion control,” Special Issue of the
IEEE Network ``Control of Best Effort Traffic”, vol. 15, no 3, pp. 28-37, May/June 2001.[FJ95] Floyd, S., and Jacobson, V., “Link-sharing and Resource Management Models for Packet Networks,” IEEE/ACM Transa
ctions on Networking, Vol. 3 No. 4, pp. 365-386, August 1995.[WTL04] H. Y. Wei, S. C. Tsao, and Y. D. Lin, “Assessing and Improving TCP Rate Shaping Over Edge Gateways,” IEEE Trans.
on Computer, vol. 53, issue 3, pp. 259-275, March 2004.[CKD02] M. Conti, M. Kumar, S. K. Das, and B. A. Shirazi, “Quality of Service Issues in Internet Web Services,” IEEE Trans. on
Computers, vol. 51, no. 6, June 2002.[PBB98] R. Pandey, J. Fritz Barnes, and R. Fritz Barnes, “Supporting Quality of Service in HTTP Servers,” Proceedings of the Se
venteenth Annual ACM Symposium on Principles of Distributed Computing, pp. 247-256, 1998.[BBK00] N. Bhatti, A. Bouch, and A. Kuchinsky, “Integrating User-Perceived Quality into Web Server Design,” Proceedings of th
e 9th International World Wide Web Conference, 2000.[CP99] L. Cherkasova and P. Phaa, “Session Based Admission Control: a Mechanism for Web QoS,” Proceedings of the Internat
ional Workshop on Quality of Service, 1999.[CCC02] V. Cardellini, E. Casalicchio, M. Colajanni, and M. Mambelli, “Enhancing a Web-Server Cluster with Quality of Service
Mechanisms,” Proceedings of IEEE International Performance Computing and Communications Conference, 2002.[CC01] E. Casalicchio and M. Colajanni, “A Client-Aware Dispatching Algorithm for Web Clusters Providing Multiple Services,
” Proceedings of the 10th International World Wide Web Conference, 2001.[LGC01] C. Li, G. Peng, K. Gopalan, and T. Chiuch, “Performance Guarantee for Cluster-Based Internet Services,” State Universit
y of Stony Brook, May 2001. [PG93] A. K. Parekh and R. G. Gallager, “A Generalized Processor Sharing Approach to Flow Control in Integrated Services Net
works: The Single-Node case,” IEEE/ACM Trans. on Networking, pp. 344-357, June 1993.[SQI06] Squid Web Proxy Cache, http://www.squid-cache.org/.[SV96] M. Shreedhar and G. Varghese, “Efficient Fair Queuing Using Deficit Round-Robin,” IEEE/ACM transactions on networ
king vol. 4, no. 3, June 1996.[GOL94] J. Golestani, “A Self-Clocked Fair Queueing Scheme for Broadband Applications,” Proceedings of the IEEE INFOCOM,
Toronto, June 1994.
51 /482007/12/12
[GVC96] P. Goyal, H. Vin, and H. Chen, “Start-Time Fair Queueing: A Scheduling Algorithm for Integrated Services Packet Switching Networks,” Proceedings of the ACM SIGCOMM, August 1996.
[BC98] P. Barford and M. Crovella, “Generating Representative Web Workloads for Network and Server Performance Evaluation,” ACM SIGMETRICS Performance Evaluation Review, vol. 26, issue 1, pp. 151-160, June 1998.
[AVA06] Avalanche, http://www.spirentcom.com/analysis/product_line.cfm?PL =32.[NFT06] The netfilter/iptables project, http://www.netfilter.org/.[TL01] Shih-Chiang Tsao and Ying-Dar Lin, “Pre-order Deficit Round Robin: A New Scheduling Algorithm for Packet-switche
d Networks,” Computer Networks, Vol. 35(2-3), pp. 287-305, 2001. [TZ05] V. Tsaoussidis and C. Zhang, “The Dynamics of Responsiveness and Smoothness in Heterogeneous Networks,” IEEE J
SEL AREA COMM, vol. 23, no. 6, pp. 1178-1189, June 2005.[MMF96] M. Mathis, J. Mahdavi, S. Floyd, and A. Romanow, “TCP Selective Acknowledgment Options, “ RFC 2018, April 1996.[F00] S. Floyd, “Congestion Control Principles,” RFC 2914, Sept. 2000.[NK04] N. Itaya and S. Kasahara, “Dynamic Parameter Adjustment for Available-bandwidth Estimation of TCP in Wired-wireles
s Networks,” Computer Communications, vol. 27 no. 10, pp. 976-988, June 2004.[G04] S. Gorinsky, "Feedback Modeling in Internet Congestion Control", in Proc. of Next Generation Teletraffic and Wired/Wi
reless Advanced Networking (NEW2AN 2004), Feb. 2004, pp. 231-234.[BP95] L. Brakmo and L. Peterson, “TCP Vegas: End to End Congestion Avoidance on a Global Internet,” IEEE Journal on Sele
cted Areas in Communication, vol 13, no. 8, Oct. 1995, pp. 1465-1480.[ZT06] C. Zhang and V. Tsaoussidis, "TCP Smoothness and Window Adjustment Strategy", IEEE Transactions on Multimedia,
vol 8, no. 3, pp. 600-609, June 2006.[KHF06] E. Kohler, M. Handley, and S. Floyd, “Designing DCCP: congestion control without reliability,” ACM SIGCOMM Comp
uter Communication Review, vol. 36, no. 4, Sept. 2006.[GTC06] L. Guo, E. Tan, S. Chen, Z. Xiao, O. Spatscheck and X. Zhang, “Delving into Internet Streaming Media Delivery: A Qua
lity and Resource Utilization Perspective,” in Proc. of ACM IMC’06, Oct. 2006. [JID04] S. Jaiswal, G. Iannaccone, C. Diot, J. Kurose, Towsley, D., “Inferring TCP connection characteristics through passive me
asurements,” IEEE INFOCOM 2004, Mar. 2004, pp. 1582-1592.[RKL01] Y. R. Yang, M. S. Kim, and S. S. Lam, “Transient behavior of TCP friendly congestion control protocols,” in Proc. of IE
EE INFOCOM, Apr. 2001, pp. 1716-1725.[JLH07] D. Wei, C. Jin, S. H. Low, and S. Hegde, “FAST TCP: Motivation, Architecture, Algorithms, Performance, “ IEEE/ACM
Transactions on Networking, vol. 14, no. 6, pp. 1246-1259, 2007.[TWH05] K. A. Tang, J. Wang, S. Hegde, and S. H. Low, “Equilibrium and Fairness of Networks Shared by TCP Reno and FAST”,
Telecommunications Systems special issue on High Speed Transport Protocols, vol. 30, no. 4, pp. 417-439, Dec. 2005.
52 /482007/12/12
[LPW03] S. H. Low, F. Paganini, J. Wang, and J. Doyle, “Linear Stability of TCP/RED and a Scalable Control,” Computer Networks Journal, vol. 43, no. 5, pp. 633-647, Dec. 2003.
[KHR02] D. Katabi, M. Handley, and C. Rohrs, “Congestion Control for High Bandwidth-Delay Product Networks,” in Proc. of ACM SIGCOMM’02, Aug. 2002, pp. 89-102.
[XSS05] Y. Xia, L. Subramanian, I. Stoica, and S. Kalyanaraman, "One More Bit Is Enough", in Proc. of ACM SIGCOMM’05, Aug. 2005
[PG93] A. K. Parekh and R. G. Gallager, “A Generalized Processor Sharing Approach to Flow Control in Integrated Services Networks: The Single-Node case,” IEEE/ACM Trans. on Networking, pp. 344-357, June 1993.
[SF03] Sally Floyd, “HighSpeed TCP for Large Congestion Windows,” RFC 3649, Dec. 2003.[YCC07] Cheng-Yuan Ho, Yi-Cheng Chan, and Yaw-Chung Chen, “WARD: A Deterministic Fluid Model,” IET Communications,
Vol. 1, Issue 4, pp. 711-717, August 2007. [YCC06] Cheng-Yuan Ho, Yi-Cheng Chan, and Yaw-Chung Chen, “Gallop-Vegas: An Enhanced Slow-Start Mechanism for TCP
Vegas,” Journal of Communications and Networks, Vol. 8, No. 3, pp. 351-359, Sept. 2006.