1. Multimedia and QoS#1 Quality of Service Support

2. Multimedia and QoS#2 QOS in IP Networks r IETF groups are working on proposals to provide QOS control in IP networks, i.e., going beyond best effort to provide some assurance for QOS r Work in Progress includes RSVP, Differentiated Services, and Integrated Services r Simple model for sharing and congestion studies:

3. Multimedia and QoS#3 Principles for QOS Guarantees r Consider a phone application at 1Mbps and an FTP application sharing a 1.5 Mbps link. m bursts of FTP can congest the router and cause audio packets to be dropped. m want to give priority to audio over FTP r PRINCIPLE 1: Marking of packets is needed for router to distinguish between different classes; and new router policy to treat packets accordingly

4. Multimedia and QoS#4 Principles for QOS Guarantees (more) r Applications misbehave (audio sends packets at a rate higher than 1Mbps assumed above); r PRINCIPLE 2: provide protection (isolation) for one class from other classes r Require Policing Mechanisms to ensure sources adhere to bandwidth requirements; Marking and Policing need to be done at the edges:

5. Multimedia and QoS#5 Principles for QOS Guarantees (more) r Alternative to Marking and Policing: allocate a set portion of bandwidth to each application flow; can lead to inefficient use of bandwidth if one of the flows does not use its allocation r PRINCIPLE 3: While providing isolation, it is desirable to use resources as efficiently as possible

6. Multimedia and QoS#6 Principles for QOS Guarantees (more) r Cannot support traffic beyond link capacity m Two phone calls each requests 1 Mbps r PRINCIPLE 4: Need a Call Admission Process; application flow declares its needs, network may block call if it cannot satisfy the needs

7. Multimedia and QoS#7 Summary

8. Multimedia and QoS#8 Scheduling And Policing Mechanisms r Scheduling: choosing the next packet for transmission m FIFO m Priority Queue m Round Robin m Weighted Fair Queuing r We had a lecture on that!

9. Multimedia and QoS#9

10. Multimedia and QoS#10 Discussion of RED r Advantages r Early drop m TCP congestion r Fairness in drops m Bursty versus non-Bursy r Disadvantages m Many additional parameters m Increasing the loss

11. Multimedia and QoS#11 Policing Mechanisms r (Long term) Average Rate m 100 packets per sec or 6000 packets per min?? crucial aspect is the interval length r Peak Rate: m e.g., 6000 p p minute Avg and 1500 p p sec Peak r (Max.) Burst Size: m Max. number of packets sent consecutively, ie over a short period of time r Units of measurement m Packets versus bits

12. Multimedia and QoS#12 Policing Mechanisms r Token Bucket mechanism, provides a means for limiting input to specified Burst Size and Average Rate. r Bucket can hold b tokens; r tokens are generated at a rate of r token/sec m unless bucket is full of tokens. r Over an interval of length t, the number of packets that are admitted is less than or equal to (r t + b).

13. Multimedia and QoS#13 Token bucket example arrivalqueuebucketsent p1 (5)-0- p2 (2)p13- p3 (1)p21p1 1p3,p2 4 5 parameters: b=5 r=3

14. Multimedia and QoS#14 Integrated Services r An architecture for providing QOS guarantees in IP networks for individual application sessions r relies on resource reservation, and routers need to maintain state info (Virtual Circuit??), maintaining records of allocated resources and responding to new Call setup requests on that basis

15. Multimedia and QoS#15 Call Admission r Session must first declare its QOS requirement and characterize the traffic it will send through the network r R-spec: defines the QOS being requested r T-spec: defines the traffic characteristics r A signaling protocol is needed to carry the R-spec and T-spec to the routers where reservation is required; r RSVP is a leading candidate for such signaling protocol

16. Multimedia and QoS#16 RSVP request (T-Spec) r A token bucket specification m bucket size, b m token rate, r m the packet is transmitted onward only if the number of tokens in the bucket is at least as large as the packet r peak rate, p m p > r r maximum packet size, M r minimum policed unit, m m All packets less than m bytes are considered to be m bytes m Reduces the overhead to process each packet m Bound the bandwidth overhead of link-level headers

17. Multimedia and QoS#17 Call Admission r Call Admission: routers will admit calls based on their R-spec and T-spec and base on the current resource allocated at the routers to other calls.

18. Multimedia and QoS#18 Integrated Services: Classes r Guaranteed QOS: this class is provided with firm bounds on queuing delay at a router; envisioned for hard real-time applications that are highly sensitive to end-to-end delay expectation and variance r Controlled Load: this class is provided a QOS closely approximating that provided by an unloaded router; envisioned for today’s IP network real- time applications which perform well in an unloaded network

19. Multimedia and QoS#19 R-spec r An indication of the QoS control service requested m Controlled-load service and Guaranteed service r For Controlled-load service m Simply a Tspec r For Guaranteed service m A Rate (R) term, the bandwidth required R  r, extra bandwidth will reduce queuing delays m A Slack (S) term The difference between the desired delay and the delay that would be achieved if rate R were used With a zero slack term, each router along the path must reserve R bandwidth A nonzero slack term offers the individual routers greater flexibility in making their local reservation Number decreased by routers on the path.

20. Multimedia and QoS#20 QoS Routing: Multiple constraints r A request specifies the desired QoS requirements m e.g., BW, Delay, Jitter, packet loss, path reliability etc r Two type of constraints: m Additive: e.g., delay m Maximum (or Minimum): e.g., Bandwidth r Task m Find a (min cost) path which satisfies the constraints m if no feasible path found, reject the connection

21. Multimedia and QoS#21 Example of QoS Routing A B D = 30, BW = 20 D = 24, BW = 55 D = 5, BW = 90 D = 3, BW = 105 D = 5, BW = 90 D = 10, BW = 90 D = 5, BW = 90 D = 7, BW = 90 D = 5, BW = 90 D = 14, BW = 90 Constraints: Delay (D) 30

22. Multimedia and QoS#22 Differentiated Services r Intended to address the following difficulties with Intserv and RSVP; r Scalability: maintaining states by routers in high speed networks is difficult sue to the very large number of flows r Flexible Service Models: Intserv has only two classes, want to provide more qualitative service classes; want to provide ‘relative’ service distinction (Platinum, Gold, Silver, …) r Simpler signaling: (than RSVP) many applications and users may only want to specify a more qualitative notion of service

23. Multimedia and QoS#23 Differentiated Services r Approach: m Only simple functions in the core, and relatively complex functions at edge routers (or hosts) m Do not define service classes, instead provides functional components with which service classes can be built

24. Multimedia and QoS#24 Edge Functions at DiffServ (DS) r At DS-capable host or first DS-capable router r Classification: edge node marks packets according to classification rules to be specified (manually by admin, or by some TBD protocol) r Traffic Conditioning: edge node may delay and then forward or may discard

25. Multimedia and QoS#25 Core Functions r Forwarding: according to “Per-Hop-Behavior” or PHB specified for the particular packet class; such PHB is strictly based on class marking (no other header fields can be used to influence PHB) r BIG ADVANTAGE: No state info to be maintained by routers!

26. Multimedia and QoS#26 Classification and Conditioning r Packet is marked in the Type of Service (TOS) in IPv4, and Traffic Class in IPv6 r 6 bits used for Differentiated Service Code Point (DSCP) and determine PHB that the packet will receive r 2 bits are currently unused

27. Multimedia and QoS#27 Classification and Conditioning r It may be desirable to limit traffic injection rate of some class; user declares traffic profile (eg, rate and burst size); traffic is metered and shaped if non-conforming

28. Multimedia and QoS#28 Forwarding (PHB) r PHB result in a different observable (measurable) forwarding performance behavior r PHB does not specify what mechanisms to use to ensure required PHB performance behavior r Examples: m Class A gets x% of outgoing link bandwidth over time intervals of a specified length m Class A packets leave first before packets from class B

29. Multimedia and QoS#29 Forwarding (PHB) r PHBs under consideration: m Expedited Forwarding: departure rate of packets from a class equals or exceeds a specified rate (logical link with a minimum guaranteed rate) m Assured Forwarding: 4 classes, each guaranteed a minimum amount of bandwidth and buffering; each with three drop preference partitions

30. Multimedia and QoS#30 Differentiated Services Issues r AF and EF are not even in a standard track yet… research ongoing r “Virtual Leased lines” and “Olympic” services are being discussed r Impact of crossing multiple ASs and routers that are not DS-capable

31. Multimedia and QoS#31 DiffServ Routers ClassifierMeterPolicerMarker DiffServ Edge Router Extract DSCP Local conditions PHB Select PHB Packet treatment DiffServ Core Router

32. Multimedia and QoS#32 IntServ vs. DiffServ IP IntServ network DiffServ network "Call blocking" approach "Prioritization" approach

33. Multimedia and QoS#33 Comparison of Intserv & Diffserv Architectures

34. Multimedia and QoS#34 Comparison of Intserv & Diffserv Architectures

35. Multimedia and QoS#35 Diffserv Theoretical Model

36. Multimedia and QoS#36 Basic Theoretical Model r Single FIFO queue. r Bounded capacity: holds up to B packets m All packets have same size r Packet Arrival: arbitrary r Packet Send: 1 packet/time unit r Actions: m Non-Preemptive model: accept or reject m Preemptive model: also preempt FIFO

37. Multimedia and QoS#37 Packet Values r Goal: m Each packet has an intrinsic value m maximize the total value of packet sent! r Cheap and expensive packets (two values): m low value of 1 and high value of  r Continuous packet values m any value in [1,  ]

38. Multimedia and QoS#38 Competitive Analysis r Analysis for online algorithms r For a given sequence S: V A (S) / V opt (S) r Competitive Ratio: MIN S {V A (S) / V opt (S)} r Worse case guarantee packets algorithm decisions

39. Multimedia and QoS#39 Non-Preemptive Policies r Fixed Partition(x) m At most xB low value and (1-x)B high value. r Flexible Partition (x) m At most xB low value and any high value. r Round Robin(x): m Like fixed partition. m send x low and (1-x) high [fractional!] m Simulate it using FIFO queue.

40. Multimedia and QoS#40 Implementing Round Robin r Implementation: m Maintain two variables: high low m If low packet arrives tests low +1 < xB IF YES ACCEPT IF NO REJECT m High packets the same m Sending: low = low –x high = high – (1-x) r Main observation: m once a packet is accepted it will be sent eventually. m Sending order not important!

41. Multimedia and QoS#41 Analysis of Round Robin r Consider the case that all packet values are 1. r Claim: m For any input sequence m The number of packet a buffer of size B/2 accepts m is at least half of a buffer of size B r Let x= ½ r Consider Low and High packets separately r RR(½) : m Accepts at least half High and half Low m Benefit at least half

42. Multimedia and QoS#42 Preemptive Policies r Greedy: m Always accept if the buffer is not full m Preempt a low value packet to accept a high one m COMPETITIVE RATIO 2 r  -Preemptive: m Drop from the head packets with total value  /  m Active queue management (AQM)

43. Multimedia and QoS#43 Preemptive Model:  1/2 -Preemptive r We consider  1/2 -Preemptive Policy r There are two packet values: 1 and  r For  =9 each high value packet preempts 3 low value packets (pro-active preemptions) low high

44. Multimedia and QoS#44  1/2 -Preemptive: Theorem r Claim 1: V A (S low )  V OPT (S low ) + 1/  1/2 V OPT (S high ) r Claim 2: V A (S high )  V OPT (S high ) + 1/  1/2 V OPT (S high ) r Theorem: V A (S)  V OPT (S) + 2/  1/2 V OPT (S)

45. Multimedia and QoS#45 Optimal Offline r Process the packet in decreasing order of value. r Accept a packet if possible. m otherwise reject r Two values: m Maximizes the number of high value packets Given a buffer of size B m Maximizes the total number of packets Using the remaining buffer space.

46. Multimedia and QoS#46 We partition the schedule to intervals: Intervals ends when the buffer is empty. Overloaded intervals: some high value packet is lost and only high value packets are scheduled. Underloaded intervals: no high value packet is lost Proof Outline: Claim 2 Overloaded Intervals time

47. Multimedia and QoS#47 Proof (Claim 1): r We show: V A (S low )  V OPT (S low ) + 1/  1/2 V A (S high ) r Low packet loss: overflow + Preemption r Low packet lost in overflow: m Opt also lost a packet. r Low packet preempted by a high packet m Value of high  m Preempted  1/2 m Value is 1/  1/2 V(high) r Recall V A (S high )  V OPT (S high )

48. Multimedia and QoS#48 We divide the HIGH packet loss into two subsets: The packets lost by OPT (easy case) The packets scheduled by OPT Proof Outline (Claim2):

49. Multimedia and QoS#49 Observation 1: When some high value packet is lost the buffer is full of high value packets B high Proof Outline (Claim 2):

50. Multimedia and QoS#50 Observation 2: If there are at least B/  1/2 high value packets in the buffer then the next packet to be scheduled is a high value packet. Proof Outline (Claim 2): high B/  1/2

51. Multimedia and QoS#51 Observation 1  The length of an overloaded interval is at least B Observation 2  An optimal offline policy could have scheduled at most B/  1/2 additional high value packets The ratio between the additional loss and the benefit of the overloaded interval is bounded by 1/  1/2 V A (S high )  V OPT (S high ) + 1/  1/2 V OPT (S high ) Proof Outline (Claim 2):

52. Multimedia and QoS#52 Lower bound (Non-Preemptive) r Scenario: m B low value packets m [maybe] B high value packets r Online: accepts xB low value m Case I: only low values Online xB Offline B m Case II: Both low and high value online xB + (1-x)  B offline  B  Competitive ratio    For large values of  we have   ½

53. Multimedia and QoS#53 Lower bound: Preemptive model r Scenario: m B low value packets m For zB time units: one high value packet arrives each time unit m [Maybe] B high value packets r Let zB be the time the Online sends the last low m (1) No more packets arrive m (2) B high value packets arrive r Online Benefit: (1) zB + z  B (2) zB +  B r Offline Benefit: (1) B + z  B (2) z  B +  B r Solving for best z gives a lower bound (about 0.8)

54. Multimedia and QoS#54 Fixed vs. Flexible Partition FixedFlexibletimeArrival event B/2 high B/2 low B high1 B/2 low B/2 high B/2 lowB/2B/2 low B/2 high B/2 lowB

55. Multimedia and QoS#55 Summary of Results: Non-preemptive Policies Always accept Round Robin Fixed Partition Flexible Partition Dynamic Flexible Partition Impossibility optimal online  =  0 ½ [1/3,0.41] 0.41 ½  =2 ½ [ 0.41,1/2] [0.41,0.56] [0.53,0.56] 2/3  =1 1 ½ 1 Two values Policies cont RR Impossibility Competitive ratio 1/(2+ ln  ) 1/(1+ ln  ) Multiple Values

56. Multimedia and QoS#56 Summary of Results: Preemptive 2 Values Policies  1/2 -Preemptive Impossibility Competitive ratio 1-2/  1/2 1-1/(2  1/2 ) Policies Greedy Better Than G Impossibility Competitive ratio ½ 1/(1.98..) 0.8 Multiple Values