Part1: Lecture 4 QoS
Last time Multi stream TCP: SCTP Multi path TCP RTP and RTCP SIP H.323 VoIP
Router architectures
Overview two key router functions: run routing algorithms/protocol (RIP, OSPF, BGP) forwarding datagrams from incoming to outgoing link forwarding tables computed, pushed to input ports routing processor routing, management control plane (software) forwarding data plane (hardware) high-seed switching fabric router input ports router output ports
Input port functions line termination link layer protocol (receive) lookup, forwarding queueing switch fabric physical layer: bit-level reception data link layer: decentralized switching: given datagram dest., lookup output port using forwarding table in input port memory ( match plus action ) goal: complete input port processing at line speed queuing: if datagrams arrive faster than forwarding rate into switch fabric
Switching fabrics transfer packet from input buffer to appropriate output buffer switching rate: rate at which packets can be transfer from inputs to outputs often measured as multiple of input/output line rate N inputs: switching rate N times line rate desirable three types of switching fabrics memory memory bus crossbar
Output ports switch fabric datagram buffer queueing link layer protocol (send) line termination buffering required when datagrams arrive from fabric faster than the transmission rate scheduling discipline chooses among queued datagrams for transmission
Remember? End-to-end delay d 1,R 1 d 2,R 2 d 3,R 3 d 4,R 4 d = ( L R i i + d i s +Q i (t))
Input port queuing fabric slower than input ports combined -> queueing may occur at input queues queueing delay and loss due to input buffer overflow! Head-of-the-Line (HOL) blocking: queued datagram at front of queue prevents others in queue from moving forward switch fabric switch fabric output port contention: only one red datagram can be transferred. lower red packet is blocked one packet time later: green packet experiences HOL blocking
Output port queueing buffering when arrival rate via switch exceeds output line speed queueing (delay) and loss due to output port buffer overflow! switch fabric switch fabric at t, packets more from input to output one packet time later
Quality of Service
One size does not fit all In the Internet IP provides a best-effort service: everybody is equal. In reality: Sensitive Delay sensitivity! Personal voice- over-ip Public Web traffic Network monitoring CEO videoconference Financial transactions Network management traffic Insensitive Personal email Business email Server backups Casual Mission criticality! Critical
Applications taxonomy Application Elastic Real-time Intolerant Tolerant Non-adaptive Adaptive Delay adaptive Rate adaptive
What is Quality of Service? The statistical performance guarantees that a network system can make in terms of throughput, delay, jitter and loss.
QoS metrics (I) Throughput: the average rate of successful message delivery over a communication channel (available) bandwidth: the net bit rate or the maximum throughput of a logical or physical communication path Goodput: the application level throughput, excluding protocol overheads
QoS metrics (II) Delay: the amount of time it takes for a packet to be transmitted end-toend across a network: Network Latency = Transmission delay + Propagation delay + Queuing delay + Processing delay Jitter: the variance in the arrival rate of packets from the same data flow " Packet loss"
A possible scenario H1 1 Mbps link R1 R2 H3 H2 1 Mbps link R1 output interface queue 1.5 Mbps link H4
How would you make sure everything works?
Marking 1 Mbps link R1 R2 1 Mbps link Principle 1 packet marking needed for router to distinguish between different classes; and new router policy to treat packets accordingly
Policing what if applications misbehave? policing: force source adherence to bandwidth allocations 1 Mbps phone R1 R2 1.5 Mbps link packet marking and policing Principle 2 provide protection (isolation) for one class from others
Efficient allocation allocating fixed (non-sharable) bandwidth to flow: inefficient use of bandwidth if flows doesn t use its allocation 1 Mbps phone R1 1 Mbps logical link R2 1.5 Mbps link 0.5 Mbps logical link Principle 3 while providing isolation, it is desirable to use resources as efficiently as possible
QoS principles 1. Router and switches can distinguish between different classes; and they can policy to treat packets accordingly Packet classification and marking!! 2. Provide protection for one class from other classes; ensure sources adhere to bandwidth requirements; to be done at the edges. Scheduling and policing!! 3. Use resources as efficiently as possible 4. Network may block application if it cannot satisfy its needs Call admission (signaling)!
Type of QoS Fine-grained approach Coarse grained approach Provide QoS to applications or specific flows Provide QoS to large classes of data or aggregated traffic Integrated Services Differentiated Services
DiffServ
Differentiated Services Learn more: Definition of the Differentiated Services Field (DS Field) in the IPv4 and IPv6 Headers RFC 2474 - Dec.1998 An Architecture for Differentiated Services RFC 2475 - Dec.1998
Test Time
Differentiated Services Differentiated Services (DiffServ) allows to give some traffic better treatment than other. Fewer drops Lower delay Lower jitter Bits in packet header indicate which packets get a better service. It deals with traffic in aggregate.
DiffServ architecture DS domain a contiguous set of DS nodes which operate with a common service provisioning policy and set of PHB groups implemented on each node. DS nodes Boundary Egress Ingress Interior
Edge and core nodes Simple functions in network core, and relatively complex functions at edge routers (or hosts. Ingress router: Police or shape traffic Set Differentiated Service Code Point (DSCP) in IP header Core router: Implement Per Hop Behavior (PHB) for each DSCP Process packets based on DSCP
PHB DiffServ supersedes the ToS field in IPv4 to make per-hop behavior (PHB) decisions about packet classification and traffic conditioning functions. PHB: results in a different observable (measurable) forwarding performance does not specify what mechanisms to use to ensure required PHB performance behavior
Learn more: Assured forwarding PHB RFC 2597 - Jun.1999 Type of service An Expedite Forwarding PHB RFC 3246 - Mar.2002 Packets are marked in the Type of Service (TOS) in IPv4, and Traffic Class in IPv6: 6 bits used for Differentiated Service Code Point (DSCP) and determine PHB that the packet will receive 2 bits are currently unused Expedited Forwarding (EF) PHB - a premium service with low-loss, low-jitter, low-delay and assured bandwidth through a DS domain. Assured forwarding (AF) PHB - superior to best effort. Does not require reservation of resources within the internet.
IP header
IP ToS 8bits ToS 0 1 2 3 4 5 6 7 IPv4 type ToS 0 1 2 3 4 5 6 7 Bits 0-2: IP-precedence defined 000 (0) - Routine 001 (1) - Priority 010 (2) - Immediate 011 (3) - Flash 100 (4) - Flash Override 101 (5) - Critical 110 (6) - Internetwork Control 111 (7) - Network Control Bits 3-6: type of service defined 0000 - All normal 1000 - Minimize delay 0100 - Maximize throughput 0010 - Maximize reliability 0001 - Minimize monetary cost
DiffServ Code Point 8bits ToS For the assured forwarding PHB: Diff Serv DS5 DS4 DS3 DS2 DS1 DS0 bits DS5, DS4 and DS3 define the class; bits DS2 and DS1 specify the drop probability; bit DS0 is always zero. Low Drop Precedence Medium Drop Precedence Class 1 Class 2 Class 3 Class 4 001010 010010 011010 100010 001100 010100 011100 100100 For the expedite forwarding PHB use codepoint 101110 High Drop Precedence 001110 010110 011110 100110
Classifiers packets Classifier Meter Marker Shaper/ dropper Classifier: select packets in a traffic stream based on the content of some portion of the packet header; the BA (Behavior Aggregate) Classifier the MF (Multi-Field) classifier
Conditioner Meter packets Classifier Marker Shaper/ dropper Meter measure the temporal properties of the stream of packets against a traffic profile! Marker set the DS field of a packet to a particular codepoint Shaper Delay (shaping) packets to bring the stream in profile Dropper discard packets to bring the stream in profile (policing)
All together Flow 1 Conditioner1 packets Classifier Flow N ConditionerN Best effort Forwarding engine
Shaping and Policing
Traffic policing parameters To correctly police traffic three main parameters to consider: Average rate The average rate expresses the flow that the source is able to sustain over an extended period of time. The interval length is crucial. Example: 100packets/sec; 6000packets/min Peak rate The maximum expected surge in traffic. Example: 6000 pkts per minute Avg and 1500 pkts per sec Variability/burst size The burstiness of a source. Max. number of pkts sent consecutively, i.e. over a short period of time
Token bucket A token bucket is a counter for the allowable number of IP bytes. It has two parameters: 1. token accumulation rate: r 2. burst tolerance: b Token rate: r (bits/s) During any period of time t the amount of data cannot exceed rt+b Bucket size: B (bytes) In Out Incoming traffic Decision point For In/Out profile
r = 50 Kbps b = 3Kb Example 3 Kb 2.1 Kb 2.2 Kb T=0; 1Kb packet arrives T=2ms: packet transmitted b= 3Kb-1kb + 2ms*50Kbps = 2.1kb T=4ms; 3Kb packet arrives 2.8 Kb 3 Kb 0.2 Kb T=16ms; packet needs to wait! T=20ms; packet transmitted T=24ms; packet transmitted
Policing! Shaping! Traffic Traffic Traffic rate Traffic rate Traffic Time Traffic Time Traffic rate Traffic rate Time Time
Pause
Scheduling
All together Flow 1 Conditioner1 packets Classifier Flow N ConditionerN Best effort Forwarding engine
FIFO First-in-first-out (FIFO) - also known as first-come-first serve (FCFS) is the traditional queuing techniques used at routers. Packets that arrive to a full buffer are either discarded, or a discard policy is used to determine which packet to discard among the arrival and those already queued xmit Arriving packets Queue Departing packets
Round robin and WRR There is a queue per flow. Queues are served in round-robin fashion. Flow1 Serve queues according to their weight Weight1 Flow2 Weight2 xmit Flow3 Weight3
Round robin problem Short packets penalized! Flow1 Flow2 xmit Flow3 Bit-round fair queuing takes into account packets sized
Fair queuing Flow1 Flow2 Bit-round robin. Compute the virtual finish time Clock ticks per bit sent Send packets in order of finish time Finish(j) F xmit Flow3 Arrive(j) F arrival time of j-th packet in flow F Length(j) F length of j-th packet in flow F Finish(j) F = max (Arrive(j) F, Finish (j-1) F ) +Length(j) F
Weighted Fair Queuing (WFQ) Generalization of Fair Queuing Provides different amounts of capacity to different flows Assign a weight, Weight F, to each flow The higher the weight the higher the bandwidth Change computation of finish time to factor in the weight Finish(j) F = max (Arrive(j) F, Finish (j-1) F ) +Length(j) F /Weight F Flow1 w 1 Flow2 w 2 Flow3 w 3 xmit
AQM Active Queue Management How to drop packets when the queue is full? 1. Tail drop All packets are identical All new packets are dropped until the queue has space xmit Arriving packets Queue Departing packets 2. Random early detection
RED It is a more proactive approach to congestion management. It monitors the average queue size and drops packets based on statistical probabilities. TH max TH min 0 Discard Discard with increasing probability P a Do not discard RED variations can accommodate QoS: WRED Weighted RED RIO RED in/out
Forwarding queues Two queues: High priority for EF traffic Low priority for AF traffic Low priority implements RED I/O DSCP? High priority queue Packets out Low priority queue RIO queue management
Packet marking at Layer2
CoS Class of Service. Achieving some more fine grained control at the Ethernet level Not the same as QoS, as it is not an end-to-end guarantee. 3 bits in the Ethernet frame indicate the PCP Priority Code Point. 802.1Q (VLAN tagging)
CoS mappings Left to implementation. IETF made some recommendations: PCP! Network priority! Traffic type! 1 0 (lowest) Background 0 1 Best Effort 2 2 Excellent Effort 3 3 Critical Applications 4 4 Video, < 100 ms latency 5 5 Voice, < 10 ms latency 6 6 Internetwork Control 7 7 (highest) Network Control
IntServ
Fine grained QoS What if you want to provide QoS not to traffic classes, but to flows?
Integrated services Integrated Services (Intserv) is an Architecture for providing QOS guarantees in IP networks for individual application sessions. Learn more: Integrated Services in the Internet Architecture: an Overview RFC 1633 - Jun.1994
Main characteristics Resource reservation: routers maintain state info of allocated resources, QoS requests Admission control: admit/deny new call setup requests Should we admit all flows (Best Effort) or refuse some to preserve good performance for the running ones (Reservation and Admission Control)?
Concerns 1. Scalability: signaling, maintaining per-flow router state difficult with large number of flows 2. Flexible Service Models: Intserv has only two classes but ofetn one wants qualitative service classes with a relative service distinction: Platinum, Gold, Silver
RSVP
Signaling protocol Resource reservation prevents congestions. To reserve resources need a signaling protocol (remember SS7). Resources in the Internet are: Link bandwidths Router buffers
RSVP Learn more: Resource ReSerVation Protocol (RSVP) - Version1 Functional Specification RFC 2205 - Sep.1997 The Resource Reservation Protocol (RSVP) is a transport layer protocol: is used by a host to request specific qualities of service from the network for particular application data streams or flows. is used by routers to deliver QoS requests to all nodes along the path(s) of the flows and to establish and maintain state to provide the requested service. RSVP-TE Traffic Engineering extension used in MPLS signaling!
RSVP messages PATH message Sent by the sender along to data path (calculated by routing protocol) Sets up the path state along the data path Receiving end node will make a reservation based on request s parameters RESV message Sent from the receiver along the reverse data path Sets up the reservation state at routers along path A receiver-oriented protocol! Once the data stream has completed, a PATHTEAR and RESVTEAR is triggered to terminate the call and release resources back to the mainstream traffic.
R1 S1 R2 S2 Internet R3 S3 PATH messages R1 S1 R2 S2 Internet R3 S3 RESV messages
Route pinning What if you have asymmetric routing? S2 S -> S1 -> S2 -> S3 -> R S S1 S3 R and R -> S3 -> S5 -> S4 > S1 -> S S4 S5 S2 S S1 S3 R Use PATH to remember the path to S! Route pinning. S4 S5
RSVP does not specify how resources are to be reserved rather: a mechanism for communicating needs determine routes packets will take that s the job of routing protocols signaling decoupled from routing interact with forwarding of packets separation of control (signaling) and data (forwarding) planes
All together Routing messages RSVP messages Routing RSVP Admission control Control plane Forwarding tables Per Flow QoS table Data plane Route lookup Classifier Scheduler
Comparison Best Effort! DiffServ! IntServ!! Service Connectivity No isolation No guarantees Per aggregate isolation Per aggregate guarantee Per flow isolation Per flow guarantee Service scope End-to-end Domain End-to-end Complexity No setup Long term setup Per flow setup Highly scalable Scalable Not scalable Scalability (nodes maintain only routing state) (edge routers maintain per aggregate state; core routers per class state) (each routers maintain per flow state)
Home reading For the test on Apr. 15 read: TMD and SONET/SDH basics Chapter 2 - Section 2.1, 2.2 and (part of) 2.3 from Optical Network Control by G. Bernstein et al.
Literature Section 4.3 What s inside a router? Chapter 17 Integrated and Differentiated Section 7.5 Network support for multimedia Services Chapter 18 Protocols for QoS support