Internet Congestion Control Speaker ： Yi-Cheng Chan ( 詹益禎 ) Assistant Professor, Department...

Internet Congestion Internet Congestion ControlControl

Speaker ： Yi-Cheng Chan (詹益禎 )Assistant Professor,

Department of Computer Science and Information Engineering,

National Changhua University of Education

2006/10/27

NCUE CSIE Y. C. Chan 2

OpeningOpening

What is the Internet? A network is a group of connected, communicating

devices. An internet is two or more networks that can

communicate with each other. The most notable internet is called the Internet.

How congestion control in the Internet? Hosts & network devices


OutlineOutline

The Internet Transmission Control Protocol - TCP Active Queue Management - AQM Admission Control Conclusions


The InternetThe Internet

Internet history 1969: Four-node ARPANET established. 1972: Vint Cerf and Bob Kahn’s Internet Project 1973: Development of TCP/IP begins. 1981: UNIX operating system includes TCP/IP. 1983: TCP/IP became the official protocol for the ARPANET. 1986: NSFNET (sponsored by the National Science

Foundation) 1990: ARPANET retired. 1991: A high-speed Internet backbone called ANSNET was

build by IBM, Merit, and MCI. 1995: Companies known as Internet Service Providers

(ISPs) started.


Internet TodayInternet Today


Growth of the InternetGrowth of the Internet

The Internet has grown tremendously. The Internet is still growing.

New Protocols New Technology Increasing use of multimedia


Internet Congestion Control ComponentsInternet Congestion Control Components

Internet

BackboneRouter

AccessRouterEnd Host

End Host

AccessRouter

Admission Control

Admission Control

Transmission Control Protocol

Transmission Control Protocol

Active Queue Management


Transmission Control Protocol - TCPTransmission Control Protocol - TCP


TCP Services (1/2)TCP Services (1/2)

Process-to-process communication

Well-known ports used by TCP


TCP Services (2/2)TCP Services (2/2)

Virtual connection Three-way handshaking Flow control

It regulates the amount of data a source can send before receiving an ACK from the destination.

Error control Checksum, ACK, retransmission time-out.

Congestion control The mechanism to control the congestion and keep the load

below the capacity.


Throughput versus Network LoadThroughput versus Network Load


Congestion Control in TCP (1/5)Congestion Control in TCP (1/5)

Congestion window Slow start: exponential increase



Congestion avoidance: additive increase



Fast retransmission



TCP congestion control policy summary



Congestion example


Evolutions of TCPEvolutions of TCP

RFC 793 (1981) A simple sliding window flow control mechanism.

Tahoe (1988) Slow start, congestion avoidance, fast retransmit.

Reno (1990) Fast recovery.

New Reno (1995) Refine the fast retransmit.


Advanced EnhancementsAdvanced Enhancements

TCP with selective acknowledgments (1996) TCP Vegas (1994) Compound TCP (2006) And …


Active Queue Management - AQMActive Queue Management - AQM

Queue management is defined as the algorithms that manage the length of queues by dropping packets when necessary or appropriate.

Why Queue Management is needed? Passive Queue Management (PQM)

Drop packets when router buffer gets full. Active Queue Management (AQM)

Employ preventive packet drop before the router buffer gets full.


Passive Queue Management (1/2)Passive Queue Management (1/2)

Two states of PQM: No packet drop 100% packet drop

Drop-Tail Drop packets from the tail of the queue. All arriving packets are dropped once the queue size

reaches a certain threshold.


Passive Queue Management (2/2)Passive Queue Management (2/2)

Drop-Front Drop the packet in the buffer with the oldest age.

Push-Out The latest buffered packet is pushed out from the

queue.


Problem with PQMProblem with PQM

Lock out Global synchronization Full queue


Active Queue ManagementActive Queue Management

Provide preventive measures to manage a buffer to eliminate problems associated with PQM.

Characteristics: Preventive random packet drop is performed before

the buffer is full The probability of preventive packet drop increases

with the increasing level of congestion. Goals:

Reduce dropped packets Support low-delay interactive services Improve the fairness


Random Early Detection (RED)Random Early Detection (RED)

A router maintains two thresholds: Minth:

Accept all packets until the queue reaches Minth

Drop packets with a linear drop probability when the queue is greater than Minth

Maxth: All packets are dropped with probability of Maxdrop when the

queue exceeds this threshold


Selection of MaxSelection of Maxdropdrop for RED for RED

Selection of Maxdrop significantly affects the performance of RED Too small: Active packet drops not enough to prevent

global synchronisation Too large: Decreases the throughput Optimal value depends on number of connections,

round trip time, etc.

Selection of an optimal value for Maxdrop remains an open issue


Calculation of the Average Queue LengthCalculation of the Average Queue Length

The average queue length controls the active packet drop. Accumulate short term congestion and trace long term

congestion.

Average queue length works as a low pass filter (LPF).

LPF/ODA


RED VariantsRED Variants

Aggregate control Modifying the calculation of the control variable and/or

drop function. Determines packet drop probability. SRED, DSRED, REM, BLUE, …

Per-flow control Configuring and setting RED’s parameters. Addresses fairness problem. FRED, FB-RED, XRED, …


Admission ControlAdmission Control

Why admission control? Expressing QoS can be done in flow

specification.

Characteristics of the Input Service Required

maximum data unit size (bytes) Token bucket rate (bytes/sec) Toke bucket size (bytes) Maximum transmission rate (bytes/sec)

Loss sensitivity (bytes) Loss interval (sec) Burst loss sensitivity (data units) Minimum delay noticed (sec) Maximum delay variation (sec) Quality of guarantee


Leaky Bucket (1/2)Leaky Bucket (1/2)

At the host-network interface, allow packets into the network at a constant rate.

Packets may be generated in a bursty manner, but after they pass through the leaky bucket, they enter the network evenly spaced.

LeakyBucket

Network

Packets from host


Leaky Bucket (2/2)Leaky Bucket (2/2)

The leaky bucket is a “traffic shaper”: It changes the characteristics of packet stream.

Traffic shaping makes traffic more manageable and more predictable.

In some cases, we may want to allow short bursts of packets to enter the network without smoothing them out.

For this purpose we use a token bucket, which is a modified leaky bucket.


Token Bucket (1/2)Token Bucket (1/2)

The bucket holds tokens instead of packets. Tokens are generated at a constant rate. When a packet arrives at the token bucket, it is

transmitted if there is a token available. Otherwise it is buffered until a token becomes available.

Tokens are buffered in a bucket with limited capacity. Bucket is full, tokens are dropped.

Data are passed to network in a relatively constant rate.

Allows for some bursty traffic from the application.


Token Bucket (2/2)Token Bucket (2/2)


Token Bucket vs. Leaky Bucket (1/2)Token Bucket vs. Leaky Bucket (1/2)

Case 1: Short burst arrivals

6543210

Arrival time at bucket

Departure time from a leaky bucketLeaky bucket rate = 1 packet / 2 time unitsLeaky bucket size = 2 packets

6543210

6543210

Departure time from a token bucketToken bucket rate = 1 tokens / 2 time unitsToken bucket size = 2 tokens(There are two tokens in the bucket at time 0.)


Token Bucket vs. Leaky Bucket (2/2)Token Bucket vs. Leaky Bucket (2/2)

Case 2: Large burst arrivals

6543210

Arrival time at bucket

Departure time from a leaky bucketLeaky bucket rate = 1 packet / 2 time unitsLeaky bucket size = 2 packets

6543210

6543210

Departure time from a token bucketToken bucket rate = 1 token / 2 time unitsToken bucket size = 2 tokens(There are two tokens in the bucket at time 0.)


ConclusionsConclusions

How congestion control in the Internet? Challenges

Increasing traffic volume Varied QoS requirements Complicated networking environment


Thank Youand

Good-bye !!


選擇題選擇題 (1/2)(1/2)

( ) 下列何者不是執行網際網路壅塞控制的元件 (1) end host (2) access router (3) backbone router (4) network link

( ) 下列何者接替了 ARPANET 成為網際網路的骨幹網路 (1) NSFNET (2) ANSNET (3) ETHERNET (4) FDDI

( ) TCP/IP 最先被內建在哪一個作業系統中 (1) Windows (2) Dos (3) Mac (4) UNIX

( ) 下列何者不是 Internet Service Provider (1) HiNet (2) SeedNet (3) AppNet (4) TaNet

( ) 下列何者不是應用層協定 (1) SNMP (2) DNS (3) ARP (4) SMTP

( ) HTTP (Hyper Text Transmission Protocol) 的 TCP well known port 是 (1) 80 (2) 23 (3) 67 (4) 53

( ) 何者不包含在 TCP congestion control policy 之中 (1) quick shift (2) congestion avoidance (3) fast retransmission (4) slow start

( ) 哪一個版本的 TCP 開始加入 fast recovery 功能 (1) Tahoe (2) Vegas (3) New Reno (4) Reno


選擇題選擇題 (2/2)(2/2)

( ) 何者不屬於 Passive Queue Management 演算法 (1) drop-tail (2) drop-front (3) drop-middle (4) push out

( ) 何者不是 Passive Queue Management 的缺點 (1) lock out (2) global synchronization (3) empty queue (4) full queue

( ) 相較於 Passive Queue Management ，何者不是 Active Queue Management 的改進目標 (1) reduce dropped packets (2) support low-delay interactive services (3) increase link utilization (4) improve the fairness

( ) 何者不是 Random Early Detection (RED) 需要的參數 (1) MAXth (2) average queue length (3) total throughput (4) MAXdrop

( ) 在 Random Early Detection (RED) 的下列改進方案中何者不屬於aggregate control (1) SRED (2) FRED (3) DSRED (4) REM

( ) 何者為 admission control 的方法之ㄧ (1) leaky bucket (2) gum bucket (3) leaf bucket (4) toy bucket

( ) 何者不是網際網路壅塞控制可見未來的主要挑戰 (1) increasing traffic volume (2) varied QoS requirements (3) complicated networking environment (4) 以上皆是

Internet Congestion Control Speaker ： Yi-Cheng Chan ( 詹益禎 ) Assistant Professor, Department...

Documents

Transcript of Internet Congestion Control Speaker ： Yi-Cheng Chan ( 詹益禎 ) Assistant Professor, Department...