What Quality of Service is About

1. What Quality of Service is About Hanoch Levy Feb 2004 Hanoch Levy, CS, TAU

2. Outline • What is Quality of Service on the Internet about • What the aims of this workshop • Structure of the course Hanoch Levy, CS, TAU

3. The Users and the service: What Communications Network serve • User A and User B • Placed in different locations • Want to pass data of some type, from one to another. • Want this to be done good/best/ASAP. • Want to do it with certain minimal quality requirements. Hanoch Levy, CS, TAU

4. The Users and the service: The user point of view • The single User is at the center of focus • User is interested in getting the quality / performance it wants • Selfish: User does not care about other users!! Hanoch Levy, CS, TAU

5. The Scarce Resource problem • Resources are (are’nt?) limited • Giving the user as much as it needs/ wants may be difficult due to limited resource problem. • Subjects where there is not resource limitation – do not face Quality of service problem. Hanoch Levy, CS, TAU

6. The Scare Resource problem: Does it exist • Are car-road resources limited? • Definitely, At least at large fractions of the day. • Are Network resources limited: • Open question • A debated question • A subjective question: (cost-wise? Indirect-cost? Damage?) Hanoch Levy, CS, TAU

7. The Scarce Resource problem: A Moving Target / Relative issue • Technology and economics are changing • Technology LEAPS: • Open bottlenecks • Move bottlenecks / create bottlenecks • Examples: • Widening a major road removes traffic jam on road and create one in the city entrance. • Creation of WEB on mid 90’s: • Makes access to WEB servers easy • Creates traffic jams on the internet. • The hot issue of today, is not that of tomorrow Hanoch Levy, CS, TAU

8. The Scarce NETWORK Resource problem: Does it exist? - revisit • Late 90’s: Telecomm boom: • Massive investment in infrastructure, especially at long distance •  Many claim: today a lot of “bandwidth in the ground” •  No resource problem for communications. • Really? • Access networks? Still may be a problem • Wireless networks? Can be a major problem. Hanoch Levy, CS, TAU

9. The Scarce NETWORK Resource problem: Does it exist? (3) • Note: • “no problem” of long distance  due to major economics investment • “wireless problem”  due to technological leap in wireless Hanoch Levy, CS, TAU

10. Communications Network and QoS :The network angle • The network serves MANY users • The network aims at providing good/Best quality to all of its users. • The network must account for the needs of all users and achieve a mechanism that can meet them. Hanoch Levy, CS, TAU

11. Communications Network and QoS :The network angle • Communications Network is like a set of car roads • Communication applications are like streams of cars. • QoS deals with how to operate these roads in order to provide the cars with good quality of service. Hanoch Levy, CS, TAU

12. How does it look like: the network from 10K feet Hanoch Levy, CS, TAU

13. The User perspective / The application perspective • Given the set of traffic rules used by the network user/ application want to: • Get the quality it wants • As good as possible? • Good above certain quality? • Pay as little as possible to get that quality. • User probably does not care about: • Other users (their quality) • Fairness to other users. Hanoch Levy, CS, TAU

14. The User perspective / The application perspective • The application: • May care or may not care about other users / their “fairness”, etc. • This depends with what perspective the application was built. • Social objective / or : • Selfishness Hanoch Levy, CS, TAU

15. The eternal race between the network and the users • There is an eternal race/game between the network and the users. • Network set up rules of operations • Users try to exploit them • Go back to 1. • (Sometimes: technology change, and goto 1) • The tighter the rules of operation the less freedom the user has. • The tighter the rules of operation the better quality is granted to the users. Hanoch Levy, CS, TAU

16. Tightness of operational rules: example1 : Car traffic • The Transport (car) system • Semi loose system • One CAN drive 200 KM/Hour • One Can cross red lights • One can drive on sideway / and/or abuse lanes •  getting better performance one’s car •  on the account of others. • Still rules are strict enough • One cannot really do it for long time • Where is the looseness: • Rules are strict • Enforcement less strict. Hanoch Levy, CS, TAU

17. Tightness of operational rules: example1 : Car traffic • Most abuse is on speed / priority at junctions • No abuse at volume (bandwidth): • No limitation on the number of cars one can buy and send into the street. • Reason: • Not really necessary • Car is so expensive – one cannot really send many cars in (bottleneck in the car supply!). • Q: what will happen if car cost drops by factor of 5? (This is what happened to communications costs!!) • Probably increase, but then flatten (hard to drive 2 cars concurrently!) Hanoch Levy, CS, TAU

18. todo Tightness of operational rules: Example2: Telephony • The Telephone system (non cellular) • Very strict system: • User has almost NO CONTROL of how application behaves • User has almost no control of the resources she gets from the network • User gets VERY GOOD QUALITY • What happens if phone cost drops by factor of 5? • Talk more… (perhaps not 5 times more). • And then flatten out (cannot talk to 10 people concurrently!!) Hanoch Levy, CS, TAU

19. todo Tightness of operational rules: Example3: Internet • The Internet system • Quite loose system (“freedom”): • Application has some freedom in the traffic it introduces to the network • The way it sends the data (later) • The user has freedom in how / how much it uses the application: • One can send as much email as one wants. • One can hit the browser button as hard as one wants. • One can download songs 24 hours a day. Hanoch Levy, CS, TAU

20. todo Tightness of operational rules: Is “freedom” (Internet) good? • The “unlimited resource” paradox: • How long can an unlimited resource remain unlimited? • In theory – perhaps long • In practice: Very few resources remained unlimited for long time. • Road? Never! • Food? Even bread hardly! • Water: almost never (cost of shipping) • Air? Hanoch Levy, CS, TAU

21. todo What happens if we get 5 times more the service • If car prices drop by factor of 5? • Probably drive more • But then flatten – cannot drive 2 cars concurrently • If telephone costs drop by factor of 5: • Probably talk more. • But then flatten -- cannot talk to 10 people concurrently • If Internet price goes down by factor of 5 (or Bandwidth goes up by factor of 5): • We can use quite a bit of it • And when we reach our limit: Probably new applications (High definition TV, 3D movies, Monitor your kid/pet/refrigerator/…. Hanoch Levy, CS, TAU

22. Objective of this workshop • To study and understand the quality of service issues of the Internet. • Understand the QoS problems • Understand the mechanisms that are used / can be used to provide QOS. • Create these mechanisms Hanoch Levy, CS, TAU

23. Methodology and contents of the course • Theoretical background – will be provided at class. • Practical experience – at the lab. • Take the user (“abuser”?) perspective: • Given the network and the network rules, client aims at maximizing its performance. • What can client do / how should client operate. Hanoch Levy, CS, TAU

24. Methodology : Project description • You have a client and your objective is to transfer (receive) X files from the network. • The files are distributed over N locations. • Some may appear in multiple locations. • The rate of downloading the files may depend on several parameters, some under your control. Hanoch Levy, CS, TAU

25. Methodology : Project description 2 • Your aim is to download the files to your best satisfaction: • As fast as possible • At lowest Bandwidth consumption(?) Hanoch Levy, CS, TAU

26. Methodology : Project description Part I • You are given both the client and the server • You aim at building a mechanism (protocol) that will transfer from server to client at maximum “efficiency” • Only minimize time • Also minimize lost resources (lost packets) Hanoch Levy, CS, TAU

27. Methodology : Project description Part II • You are given the client only • K Servers are given and they operate according to their protocol (FTP) • Want to download the files efficiently from the servers • Only minimize time • Also minimize lost resources (lost packets) Hanoch Levy, CS, TAU

28. todo QOS Problems of Interest For network designer • Traffic classification and characterization • Properties of traffic (many to choose from) • Requirements of the application / traffic • Requirements of the system • Impact on the system Hanoch Levy, CS, TAU

29. todo QOS Problems of interest (cont): 2. Policing and shaping • Monitor traffic for obeying the rules • Location: typically at network entrance 3. Node (“Intersection”) design: • Create fast intersections • Introduce mechanisms of prioritization into the intersections • Guarantee QoS to a traffic stream despite interference of other streams (fair queueing) • Location: In the nodes Hanoch Levy, CS, TAU

30. todo 18/3 QOS Problem of interest (cont) • 4. Do not overflow your nodes (intersections) – estimate node capacity (Call Admission Control) • 5. Efficient navigation of traffic (Routing) while obeying QoS • 6. Managing your traffic: Virtual paths (transform your cars into trains…) Hanoch Levy, CS, TAU

31. todo QOS Problem of interest (cont) • 7.Coordinate through network nodes (reservations): Traffic engineering. • 8. Traffic characterization. Hanoch Levy, CS, TAU