DigiComm II-1 TAPAS EB/IAB Meeting Newcastle, 5/9/02 Real inter-domain paths are unlikely to offer explicit SLA although limited local SLAs are avail.

Slides:



Advertisements
Similar presentations
Congestion Control and Fairness Models Nick Feamster CS 4251 Computer Networking II Spring 2008.
Advertisements

7. 7 Chapter 13 Transmission Control Protocol (TCP) Retransmission and Time-Out.
Streaming Video over the Internet
1 School of Computing Science Simon Fraser University CMPT 771/471: Internet Architecture & Protocols TCP-Friendly Transport Protocols.
Internetworking II: MPLS, Security, and Traffic Engineering
William Stallings Data and Computer Communications 7 th Edition Chapter 13 Congestion in Data Networks.
CS 408 Computer Networks Congestion Control (from Chapter 05)
Computer Networks: TCP Congestion Control 1 TCP Congestion Control Lecture material taken from “Computer Networks A Systems Approach”, Fourth Edition,Peterson.
XCP: Congestion Control for High Bandwidth-Delay Product Network Dina Katabi, Mark Handley and Charlie Rohrs Presented by Ao-Jan Su.
Streaming Video over the Internet: Approaches and Directions Dapeng Wu, Yiwei Thomas Hou et al. Presented by: Abhishek Gupta
Multiple constraints QoS Routing Given: - a (real time) connection request with specified QoS requirements (e.g., Bdw, Delay, Jitter, packet loss, path.
Computer Networks: TCP Congestion Control 1 TCP Congestion Control Lecture material taken from “Computer Networks A Systems Approach”, Third Ed.,Peterson.
Teknik Routing Pertemuan 20 Matakuliah: H0484/Jaringan Komputer Tahun: 2007.
1 Internet Networking Spring 2003 Tutorial 11 Explicit Congestion Notification (RFC 3168)
Adaptive Delay Aware Error Control for Internet telephony Catherine Boutremans Jean-Yves Le Boudec IP Telephony Workshop’2001 Institute for computer Communication.
Multimedia Applications r Multimedia requirements r Streaming r Phone over IP r Recovering from Jitter and Loss r RTP r Diff-serv, Int-serv, RSVP.
ACN: Congestion Control1 Congestion Control and Resource Allocation.
Semester Copyright USM EEE449 Computer Networks Congestion En. Mohd Nazri Mahmud MPhil (Cambridge, UK) BEng (Essex, UK) Room.
1 Spring Semester 2007, Dept. of Computer Science, Technion Internet Networking recitation #8 Explicit Congestion Notification (RFC 3168) Limited Transmit.
Rethinking Internet Traffic Management: From Multiple Decompositions to a Practical Protocol Jiayue He Princeton University Joint work with Martin Suchara,
CS 268: Future Internet Architectures Ion Stoica May 6, 2003.
Computer Networks Transport Layer. Topics F Introduction  F Connection Issues F TCP.
COS 420 Day 17. Agenda Finished Grading Individualized Projects Very large disparity in student grading No two students had same ranking for other students.
Reliable Transport Layers in Wireless Networks Mark Perillo Electrical and Computer Engineering.
Advanced Computer Networks: TCP Congestion Control 1 TCP Congestion Control Lecture material taken from “Computer Networks A Systems Approach”, Fourth.
CMPE 257 Spring CMPE 257: Wireless and Mobile Networking Spring 2005 E2E Protocols (point-to-point)
CS332 Ch. 28 Spring 2014 Victor Norman. Access delay vs. Queuing Delay Q: What is the difference between access delay and queuing delay? A: I think the.
Item 2005 L A Rønningen. Reservation Model Pessimistic or Optimistic Approach 1-N Senders and 1-M Receivers Sender-oriented or Receiver-oriented Immediate.
1 National Research Council - Pisa - Italy Marco Conti Italian National Research Council (CNR) IIT Institute MobileMAN Architecture and Protocols 2nd MobileMAN.
1 Flow Identification Assume you want to guarantee some type of quality of service (minimum bandwidth, maximum end-to-end delay) to a user Before you do.
Data Transmission Over Wireless Links Fan Yang
1 PWE3 Architecture PWE3 IETF March 2003 Stewart Bryant.
Chapter 8: Internet Operation. Network Classes Class A: Few networks, each with many hosts All addresses begin with binary 0 Class B: Medium networks,
Congestion control for Multipath TCP (MPTCP) Damon Wischik Costin Raiciu Adam Greenhalgh Mark Handley THE ROYAL SOCIETY.
Requirements for Simulation and Modeling Tools Sally Floyd NSF Workshop August 2005.
Networking Fundamentals. Basics Network – collection of nodes and links that cooperate for communication Nodes – computer systems –Internal (routers,
TCP with Variance Control for Multihop IEEE Wireless Networks Jiwei Chen, Mario Gerla, Yeng-zhong Lee.
DCCP: Issues From the Mailing List Sally Floyd, Eddie Kohler, Mark Handley, et al. DCCP WG March 4, 2004.
Page 1 The department of Information & Communications Engineering Dong-uk, kim A Survey of Packet Loss Recovery Techniques for Streaming.
Thoughts on the Evolution of TCP in the Internet (version 2) Sally Floyd ICIR Wednesday Lunch March 17,
Jennifer Rexford Fall 2014 (TTh 3:00-4:20 in CS 105) COS 561: Advanced Computer Networks TCP.
Low Latency Adaptive Streaming over TCP Authors Ashvin Goel Charles Krasic Jonathan Walpole Presented By Sudeep Rege Sachin Edlabadkar.
Teknik Routing Pertemuan 10 Matakuliah: H0524/Jaringan Komputer Tahun: 2009.
Users, Pricing and Resource Reservation: Managing Expectations. Jon Crowcroft,
Ασύρματες και Κινητές Επικοινωνίες Ενότητα # 11: Mobile Transport Layer Διδάσκων: Βασίλειος Σύρης Τμήμα: Πληροφορικής.
UCL TAPAS meeting 1 TAPAS Cambridge Computer lab Jon Crowcroft 2nd Feb, 2004.
TCP continued. Discussion – TCP Throughput TCP will most likely generate the saw tooth type of traffic. – A rough estimate is that the congestion window.
Optimization Problems in Wireless Coding Networks Alex Sprintson Computer Engineering Group Department of Electrical and Computer Engineering.
1 Advanced Transport Protocol Design Nguyen Multimedia Communications Laboratory March 23, 2005.
© Janice Regan, CMPT 128, CMPT 371 Data Communications and Networking Congestion Control 0.
Uni Innsbruck Informatik th IETF, PMTUD WG: Path MTU Discovery Using Options draft-welzl-pmtud-options-01.txt Michael Welzl
Chapter 10 Congestion Control in Data Networks and Internets 1 Chapter 10 Congestion Control in Data Networks and Internets.
U Innsbruck Informatik - 1 Specification of a Network Adaptation Layer for the Grid GGF7 presentation Michael Welzl University.
Internet Networking recitation #9
Congestion Control, Internet transport protocols: udp
CSE679: Multimedia and Networking
IP - The Internet Protocol
Congestion Control (from Chapter 05)
Internet Networking recitation #10
Congestion Control (from Chapter 05)
IP - The Internet Protocol
Congestion Control (from Chapter 05)
Congestion Control (from Chapter 05)
Congestion Control (from Chapter 05)
Congestion Control (from Chapter 05)
Congestion Control (from Chapter 05)
IP - The Internet Protocol
DCCP: Issues From the Mailing List
Lecture 6, Computer Networks (198:552)
Impact of transmission errors on TCP performance
Presentation transcript:

DigiComm II-1 TAPAS EB/IAB Meeting Newcastle, 5/9/02 Real inter-domain paths are unlikely to offer explicit SLA although limited local SLAs are avail. Instead, model/measure inter-domain implicit SLA Assuming Internet Exchanges are not under- provisioned, can expect SLA between 2 ISPs to be Worst of 2 (or in general, n) ISPs Net SLAs current: 90%ile delay, availability, NOT b/w though, some offer 90%ile loss too…

DigiComm II-2 Given Network SLA Above Transport Service, we still need an SLA Lets say it specifies Throughput, delay, delay jitter, residual loss Availability So the IP level has some packet loss probability – however this is NOT the important factor for the transport protocol in today’s network – the congestion level (provisioning) is – see Gibbens/Kelly/Macfadyen et al

DigiComm II-3 Transport SLA All transport protocols today are being designed to be “TCP friendly” – this means that their target throughput (transmission rate) is typically equation driven (this applies to multicast too): Rate = 2/3 * mtu / rtt * sqrt(p) Mtu = packet size Rtt = round trip time (can assume roughly 2*delay) P = packet loss probability Net is engineered for expected set of source rates (assuming TCP mix of mice &elephants, and long term balanced matrix of source/destination pairs) So for given bottlenecks, router incurs loss to cause share

DigiComm II-4 Loss Concealment Cost (this is not an option!) We have 2 transport techniques for loss concealment (whether random noise induced, or much more commonly, congestive packet loss – note that congestive loss doesn’t mean congestion (c.f. fast retransmit) _ only persistent loss does….occasional loss is a rate feedback mechanism Retransmission (TCP, PGM) Forward erasure/error correction (ALC, and sometimes PGM)

DigiComm II-5 Costs of Loss Concealment Retransmitted packet still has independent loss probability – hence expected mean delay for packets is (assuming binomial back-off for subsequent retransmits!): E(d_) = Sum over I to infinity of rtt*2^I * (1-p)^I Luckily for us, this converges for small p! Can compute delay variance similarly…. FEC has no delay penalty, but takes a fixed delay at source to add redundancy and takes a percentage hit from the network overhead of transmitting, so its nice if you want good delay properties E(txput) = 1 + p + epsilon

DigiComm II-6 Factors for API to transport TCP masks loss (almost for ever, at least 4 minutes!) PGM, or DCCP don’t (or SCTP) and have a well specified interface for signalling the inability to recover from loss…..and can allow this to be time bounded (I.e. to “fail” a transaction…for example) Some work we need to do: Look at distribution of rates and loss in multicast! Look at solving equations above for multicast and for slow start and for non-independent packet loss (take some measurements!) Maxemchuk’s protocol does better but need external clock synch

DigiComm II-7 Conclusions/Consequences Need distributed algorithm spec for delay and delay variance and residual loss acceptability Can then, given network SLA, chose a transport protocol and a configuration, and derive Transport SLA to meet this Has a cost Can be done on a pure best effort (but provisioned) network If network offers differentiation, can choose to trade-off IP SLA cost (e.g. $!) and Transport SLA costs (delay for loss concealment, or excess load from FEC) to meet Application requirements Of course if we had global deployment of ECN…

DigiComm II-8 Pragmatic TAPAS Choices No IP multicast No IP QoS Yes, provisioning (ie delay and loss %ile SLAs) Yes, new transport work Maybe consider partial ECN deployment…

DigiComm II-9 Cambridge TAPAS Research Topics Solve Equations above Relate to network calculus and to predicate routing or similar tools for solving for Transport API SLA for a set of users, on a set of ISPs Re-do for multicast (subject to selection of appropriate multicast transport!) Expose transport APIng SLAng to higher level Look at monitoring architecture and interface!