Analysis of Internet Transport Service Analysis of Internet - - PowerPoint PPT Presentation

▶
analysis of internet transport service analysis of
SMART_READER_LITE
LIVE PREVIEW

Analysis of Internet Transport Service Analysis of Internet - - PowerPoint PPT Presentation

Jan 04, 2023 224 likes •429 views

University of Helsinki - Department of Computer Science Analysis of Internet Transport Service Analysis of Internet Transport Service Performance Performance with Active Queue Management with Active Queue Management in a QoS- -enabled

slide-1
SLIDE 1

Analysis of Internet Transport Service Analysis of Internet Transport Service Performance Performance with Active Queue Management with Active Queue Management in a QoS in a QoS-

  • enabled Network

enabled Network

University of Helsinki - Department of Computer Science Oriana Riva Oriana Riva

  • riana.riva@cs.Helsinki.FI
slide-2
SLIDE 2

12.6.2003 University of Helsinki - Department of Computer Science 2

Contents Contents

TCP Congestion Control in the Internet Router queue management algorithms

– Tail Drop discipline – Active Queue Management (AQM)

  • Goals
  • Random Early Detection (RED)

Test Arrangements

– Test Environment – Test Cases

Test Results and Analysis Conclusions

slide-3
SLIDE 3

12.6.2003 University of Helsinki - Department of Computer Science 3

TCP Congestion Control in the Internet TCP Congestion Control in the Internet

Internet has grown and TCP congestion control mechanisms are not sufficient to provide good service in all circumstances – Wired vs. Wireless – The Internet is evolving to support QoS Problem: There is a limit to how much control can be accomplished from the edges of the network A possible solution is an advanced form of router queue management such as Active Queue Management (RFC 2309) Current router algorithm employed in the Internet: Tail Drop

slide-4
SLIDE 4

12.6.2003 University of Helsinki - Department of Computer Science 4

Tail Drop buffer Tail Drop buffer

Principle:

– fix the maximum length for each queue, accept packets for the queue until the maximum length is reached, then drop subsequent incoming packets

Advantages

– easy to implement

Disadvantages

– Full Queues

  • Discriminate against bursty traffic and cause multiple

packets to be dropped

  • Synchronization of sources

– Lock-Out: a single connection or few flows manage to monopolize the available queue resources

slide-5
SLIDE 5

12.6.2003 University of Helsinki - Department of Computer Science 5

Active Queue Management (AQM) Active Queue Management (AQM)

Based on a proactive approach: – drop packets before buffer becomes full – use the average queue length as congestion indicator Can provide the following advantages for responsive flows – Control the average queuing delay and reduce end-to-end latency – Guarantee more equity in resources utilization (fairness) – Reduce the number of packets dropped in routers – Provide greater capacity to absorb bursts without dropping packets – Avoid global synchronization of sources – Avoid Lock-Out behaviour The most famous example of AQM is the Random Early Detection (RED) algorithm

slide-6
SLIDE 6

12.6.2003 University of Helsinki - Department of Computer Science 6

RED RED -

  • Algorithm

Algorithm

Calculate the average queue size (avg) by a low pass filtering with weight wq, 0<wq<1

qist=instantaneous queue length

If avg<minth do nothing If avg>maxth drop the packet Else drop (or mark) packet with probability:

count =number of unmarked packets since the last time a packet was dropped or since avg exceeded minth

ist q n q n

q w )avg w ( avg + − =

+

1

1

count·p') p'/( p ) ) (avg p'

th th th p

− = − − = 1 min /(max min max

p=1 avg minth maxth

Dropping/Marking Probability, p'

slide-7
SLIDE 7

12.6.2003 University of Helsinki - Department of Computer Science 7

Test Arrangements Test Arrangements

Simple scenario Simple scenario (

(Each scenario is tested with 20 replications)

– PC with Linux Operating System (kernel version 2.4.17) – Standard Linux TCP Implementation with SACK option (Timestamps: OFF) – The Intermediate Router employs Tail Drop with a huge buffer size – The Last-hop Router employs Tail Drop or RED without ECN – RED Parameters: buffer size=20kbytes, minth=5kbytes, maxth=18kbytes, maxp=0.1 – Metrics: Elapsed Time, Num Retx Packets, Fairness, etc..

Intermediate Router Last-hop Router

100Mbit/s 10Mbit/s

Source Destination

UDP

140 kbit/s

TCP

Wired Network Wireless Network bottleneck

slide-8
SLIDE 8

12.6.2003 University of Helsinki - Department of Computer Science 8

Test Arrangements Test Arrangements -

  • Test Environment

Test Environment

The last-hop router implements a DiffServ (DS) PHB using HTB (Hierarchical Token Bucket) packet scheduler with two service classes Each service class employs Tail Drop or RED Queuing allocates bandwidth and buffer space:

– Bandwidth: which packet to serve next (scheduling) – Buffer Space: which packet to drop next (buffer management) Buffer Management Buffer Management

Drop

Tail Drop/RED

Scheduling Scheduling

HTB

Traffic Classes Class

TCP

Class

UDP

Traffic Sources

UDP TCP TCP TCP UDP UDP

slide-9
SLIDE 9

12.6.2003 University of Helsinki - Department of Computer Science 9

Test Arrangements Test Arrangements -

  • Test Case 1

Test Case 1

bottleneck 140kbit/s TCP 1 TCP 2 TCP 3

R

TCP 1 TCP 2 TCP 3

R

data sent = 292kbytes data sent= 46.7kbytes

Workload: 3 competing TCP flows

1 service class – bottleneck bandwidth =140kbit/s 2 TCP connections start simultaneously (data traffic sent=292kbytes) The third connection starts with a delay varying from 0 up to 7 seconds (data traffic sent= 46.7kbytes)

slide-10
SLIDE 10

12.6.2003 University of Helsinki - Department of Computer Science 10

Test Arrangements Test Arrangements -

  • Test Case 2

Test Case 2

Workload: TCP and UDP traffic with DiffServ

The UDP traffic starts later and has different duration

  • 1. The UDP traffic lasts for the entire duration of the TCP transfer
  • 2. The UDP traffic ends before the TCP connection ends

Effects derived from the introduction of service differentiation:

  • 1. Unique service class for TCP and UDP traffic (available

bandwidth=140kbit/s)

  • 2. Separate service classes for TCP traffic and higher-priority

UDP traffic (class bandwidth=70kbit/s)

bottleneck (140kbit/s) TCP UDP

R

TCP UDP

R

DS

70/140kbit/s 70/100kbit/s

slide-11
SLIDE 11

12.6.2003 University of Helsinki - Department of Computer Science 11

Results Results – – Test Case 1 Test Case 1

3 competing TCP connections Results of the third connection for different starting points

Stable results and improved performance when RED is employed Tail Drop – Scenario after 2 seconds: the elapsed time doubles the RED’s one!

Elapsed Time

7 9 11 13 15 17 19 21 23 25 2 4 5 5,5

starting time of the third flow (sec) Elaps.Time (sec)

RED Tail Drop

Three-way Handshake Phase

1 2 3 4 5

2 4 5 5,5

starting time of the third flow (sec) Hand.phase (sec)

RED Tail Drop

slide-12
SLIDE 12

12.6.2003 University of Helsinki - Department of Computer Science 12

Results Results – – Test Case 1 Test Case 1

Tail Drop

5 10 15 20 25 30 35 40 45 2 4 5 5,5

starting time of the third flow(sec)

Elapsed Time (sec)

Min 25% Percentile Median 75% Percentile Max

5 1 1 5 20 25 30 35 40 45 2 4 5 5,5

starting time of the third flow (sec)

Elapsed Time (sec)

Min 25% Percentile Median 75% Percentile Max

Wide variation range Limited variation range RED

3 competing TCP connections – Elapsed Time of the 3rd flow

slide-13
SLIDE 13

12.6.2003 University of Helsinki - Department of Computer Science 13

Test Case 1 Test Case 1 -

  • Lock

Lock-

  • Out of Tail Drop

Out of Tail Drop

The exponential retransmit back-off rule doubles the retransmit timer when a retransmitted packet is lost Tail Drop: example from the scenario after 2 seconds The initial flows monopolize most of the bandwidth (phenomenon of Lock-Out)

slide-14
SLIDE 14

12.6.2003 University of Helsinki - Department of Computer Science 14

Results Results -

  • Test Case 1

Test Case 1

3 competing TCP flows Third flow’s data load is 4/25 of the initial flows’ data load

Tail Drop - Elapsed Time

10 20 30 40

2 4 5 5,5

starting time of the third flow (sec)

Elaps.Time(sec)

fastest flow slowest flow 3rd flow

RED - Elapsed Time

10 20 30 40

2 4 5 5,5

starting time of the third flow (sec)

  • Elaps. Time (sec)

fastest flow slowest flow 3rd flow

Tail Drop - Retx Packets

5 1 1 5

2 4 5 5,5

starting time of the third flow (sec)

Num Retx Pkts

fastest flow slowest flow 3rd flow

RED - Retx Packets

5 10 15

2 4 5 5,5

starting time of the third flow (sec)

Num Retx Pkts

fastest flow slowest flow 3rd flow

RED guarantees to the third flow improved and stable performance RED drops packets in proportion to the amount of bandwidth the flow uses on the output link

slide-15
SLIDE 15

12.6.2003 University of Helsinki - Department of Computer Science 15

Results Results -

  • Test Case 2

Test Case 2

Single TCP flow competing with UDP traffic lasting 5sec

1 service class : – Tail Drop: unstable results for both TCP and UDP traffic – RED: performance close to the results with 2 service classes 2 service classes (Higher priority UDP traffic) – Tail Drop and RED achieve the same performance

Num of Received UDP Pkts

50 100 150 200 250 300 2 7 1 2 1 6 18

starting time of the UDP flow (sec)

Num of Recv Pkts

Tail Drop RED Tail Drop + DS RED + DS

Elapsed Time of the TCP flow

23 23,5 24 24,5 25 25,5 26 2 7 1 2 16 18

starting time of the UDP flow (sec)

El.Time (sec)

Tail Drop RED Tail Drop + DS RED + DS

slide-16
SLIDE 16

12.6.2003 University of Helsinki - Department of Computer Science 16

Test Case 2 Test Case 2 -

  • Unfairness of Tail Drop

Unfairness of Tail Drop

TCP traffic competing with UDP traffic of equal priority The performance changes by varying the starting point

  • f the UDP traffic

TCP El. Time=25.5s UDP Rec Pkts=181 TCP El. Time=23.7s UDP Rec Pkts=127 TCP El. Time=24.5s UDP Rec Pkts=224

UDP flow starts

slide-17
SLIDE 17

12.6.2003 University of Helsinki - Department of Computer Science 17

Results Results – – Test Case 2 Test Case 2

Time Sequence Graph of Traffic TCP competing with higher- priority UDP traffic starting after 0.5s Tail Drop: el. time=8.3s Timeout expiration Start to drop packets late Discard consecutive packets RED: el. time=4.9s 3 DUPACKs Start to drop packets early Discard randomly + SACK

  • ption
slide-18
SLIDE 18

12.6.2003 University of Helsinki - Department of Computer Science 18

Results Results – – Test Case 2 Test Case 2

1 TCP flow + UDP traffic (starting after 3 seconds) in presence

  • f service differentiation

Tail Drop: el.time=35.9s; rtx pkts=1 RED: el.time=35,8s; rtx pkts=7

The dropped packets are evenly distributed over a wide range The Recovery phase with Tail Drop is roughly the double of the RED’s one

slide-19
SLIDE 19

12.6.2003 University of Helsinki - Department of Computer Science 19

Conclusions Conclusions

RED fairness

– Fair sharing of resources – Avoid Lock-Out – Stable results

RED drops more packets than Tail Drop, BUT

– By starting to drop packets early it manages to recover quickly from congestion and provides improved performance – It drops packets from each flow in proportion to the amount of bandwidth the flow uses on the output link

slide-20
SLIDE 20

12.6.2003 University of Helsinki - Department of Computer Science 20

Future Work Future Work

Emulate various wireless link characteristics on the last-hop link Add background traffic Advanced use of DiffServ with different traffic mixes Use of TCP variants Use of ECN (Explicit Congestion Notification) in combination with RED Tuning of RED parameters Wired Network

Destination

DiffServ Router DiffServ Router

Source

background traffic TCP/UDP

Seawind

Wireless Network