DOC PREVIEW
CCSU CS 490 - Computer Networking - Congestion Control

This preview shows page 1-2-3 out of 10 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 10 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 10 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 10 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 10 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

1 Computer)Networking)Conges1on)Control)Principles)of)Conges1on)Control)Conges1on:)• informally:)“too)many)sources)sending)too)much)data)too)fast)for)network)to)handle”)• different)from)flow)control!)• manifesta1ons:)– lost)packets)(buffer)overflow)at)routers))– long)delays)(queueing)in)router)buffers))• a)topE10)problem!)Conges1on)Control2 Causes/costs)of)conges1on:)scenario)1))• two)senders,)two)receivers)• one)router,)infinite)buffers))• no)retransmission)• large)delays)when)congested)• maximum)achievable)throughput)Conges1on)Control unlimited shared output link buffers Host A λin : original data Host B λout Causes/costs)of)conges1on:)scenario)2))• one)router,)finite)buffers))• sender)retransmission)of)lost)packet)Conges1on)Control finite shared output link buffers Host A λin : original data Host B λout λ'in : original data, plus retransmitted data3 Causes/costs)of)conges1on:)scenario)2))• always:)))))))))))))))))))))))))(goodput))• “perfect”)retransmission)only)when)loss:)• retransmission)of)delayed)(not)lost))packet)makes)))))))))larger)(than)perfect)case))for)same)Conges1on)Control λ"in λ"out = λ"in λ"out > λ"in λ"out “costs”)of)conges1on:))• more)work)(retrans))for)given)“goodput”)• unneeded)retransmissions:)link)carries)mul1ple)copies)of)pkt)R/2 R/2 λin λout b. R/2 R/2 λin λout a. R/2 R/2 λin λout c. R/4 R/3 Causes/costs)of)conges1on:)scenario)3))• four)senders)• mul1hop)paths)• 1meout/retransmit)Conges1on)Control λ"in Q:)what)happens)as))))))and)))))))))))increase)?)λ"in finite shared output link buffers Host A λin : original data Host B λout λ'in : original data, plus retransmitted data4 Causes/costs)of)conges1on:)scenario)3))Conges1on)Control another)“cost”)of)conges1on:))• when)packet)dropped,)any)“upstream)transmission)capacity)used)for)that)packet)was)wasted!)Host A Host B λoutApproaches)towards)conges1on)control)endEend)conges1on)control:)• no)explicit)feedback)from)network)• conges1on)inferred)from)endEsystem)observed)loss,)delay)• approach)taken)by)TCP)networkEassisted)conges1on)control:)• routers)provide)feedback)to)end)systems)– single)bit)indica1ng)conges1on)(SNA,)DECbit,)TCP/IP)ECN,)ATM))– explicit)rate)sender)should)send)at)Conges1on)Control two)broad)approaches)towards)conges1on)control:)5 TCP)conges1on)control:)Conges1on)Control ❒ goal:..TCP)sender)should)transmit)as)fast)as)possible,)but)without)conges1ng)network)❍ Q:)how)to)find)rate)just)below)conges1on)level)❒ decentralized:)each)TCP)sender)sets)its)own)rate,)based)on)implicit)feedback:))❍ ACK:)segment)received)(a)good)thing!),)network)not)congested,)so)increase)sending)rate)❍ lost.segment:)assume)loss)due)to)congested)network,)so)decrease)sending)rate)TCP)conges1on)control:)bandwidth)probing)Conges1on)Control ❒ “probing)for)bandwidth”:)increase)transmis sion )rate)on)receipt)of)ACK,)un1l)eventually)loss)occurs,)then)decrease)transmission)rate))❍ con1nue)to)increase)on)ACK,)decrease)on)loss)(since)available)bandwidth)is)changing,)depending)on)other)connec1ons)in)network)))ACKs being received, so increase rate X X X X X loss, so decrease rate sending rate time ❒ Q:)how)fast)to)increase/decrease?)❍ details)to)follow)TCP’s “sawtooth” behavior6 TCP)Conges1on)Control:)details)• sender)limits)rate)by)limi1ng)number)of)unACKed)bytes)“in)pipeline”:)– cwnd: differs)from)rwnd (how,)why?))• rwnd:)unused)buffer)space)– sender)limited)by min(cwnd,rwnd) • roughly,)• cwnd is)dynamic,)func1on)of)perceived)network)conges1on)Conges1on)Control rate = cwnd RTT bytes/sec LastByteSent-LastByteAcked ≤ cwnd cwnd!bytes RTT ACK(s) TCP)Conges1on)Control:))more)details)segment)loss)event:)reducing)cwnd!• 1meout:)no)response)from)receiver)– cut)cwnd to)1)• 3)duplicate)ACKs:)at)least)some)segments)geYng)through)(recall)fast)retransmit))– cut)cwnd)in)half,)less)aggressively)than)on)1meout)Conges1on)Control ACK)received:)increase)cwnd%❒ slowstart)phase:))❍ increase)exponen1ally)fast)(despite)name))at)connec1on)start,)or)following)1meout)❒ conges1on)avoidance:))❍ increase)linearly)7 TCP)Slow)Start)• when)connec1on)begins,)cwnd)=)1)MSS)– example:)MSS)=)500)bytes)&)RTT)=)200)msec)– ini1al)rate)=)20)kbps)• available)bandwidth)may)be)>>)MSS/RTT)– desirable)to)quickly)ramp)up)to)respectable)rate)• increase)rate)exponen1ally)un1l)first)loss)event)or)when)threshold)reached)– double)cwnd)every)RTT)– done)by)incremen1ng)cwnd)by)1)for)every)ACK)received)Conges1on)Control Host A one segment RTT Host B time two segments four segments TCP:)conges1on)avoidance)• when)cwnd > ssthresh)grow)cwnd)linearly)– increase)cwnd by)1)MSS)per)RTT))– approach)possible)conges1on)slower)than)in)slowstart)– implementa1on:)cwnd = cwnd + MSS/cwnd)for)each)ACK)received)Conges1on)Control ❒ ACKs:)increase)cwnd%by)1)MSS)per)RTT:)addi1ve)increase)❒ loss:)cut)cwnd)in)half)(nonE1meoutEdetected)loss)):)mul1plica1ve)decrease)AIMD)AIMD:)Addi1ve)Increase)Mul1plica1ve)Decrease)8 Popular)“flavors”)of)TCP)Conges1on)Control ssthresh ssthresh TCPTahoeTCPRenoTransmissionroundcwnd windowsize(insegments)Summary:)TCP)Conges1on)Control)• when)cwnd < ssthresh,)sender)in)slowEstart)phase,)window)grows)exponen1ally.)• when)cwnd >= ssthresh,)sender)is)in)conges1onEavoidance)phase,)window)grows)linearly.)• when)triple)duplicate)ACK)occurs,)ssthresh set)to)cwnd/2, cwnd set)to)~)ssthresh • when)1meout)occurs,)ssthresh)set)to)cwnd/2, cwnd)set)to)1)MSS.))Conges1on)Control9 TCP)Fai rness)fairness)goal:)if)K)TCP)sessions)share)same)bobleneck)link)of)bandwidth)R,)each)should)have)average)rate)of)R/K)Conges1on)Control TCP connection 1 bottleneck router capacity R TCP connection 2 Why)is)TCP)fair?)Two)compe1ng)sessions:)• Addi1ve)increase)gives)slope)of)1,)as)throughout)increases)• mul1plica1ve)decrease)decreases)throughput)propor1onally))Conges1on)Control R R equal bandwidth share Connection 1 throughput Connection 2 throughput congestion avoidance: additive increase loss: decrease window by factor of 2 congestion avoidance: additive increase loss: decrease window by factor of 210 Fairness)(more))Fairness)and)UDP)•


View Full Document
Download Computer Networking - Congestion Control
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Computer Networking - Congestion Control and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Computer Networking - Congestion Control 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?