Internet Quality of Service

Feedback & Grades More preparation by both presenting and non-presenting students (for questions & discussions) More examples to explain concepts Grades for presentation will be sent out by Friday

Motivation for QoS? Real-time applications Convergence in the Internet Users willing to pay for better service

QoS Parameters Bandwidth Delay Delay jitter Loss Goodput

Types of QoS Absolute Relative 100Kbps, 5ms delay bound, 2% loss rate Olympic model (gold, silver, bronze) Gold better than silver, etc. Gold gets k times bandwidth as silver

Approaches Integrated Services Model Differentiated Services Model Intserv Can provide per flow QoS Problem? Differentiated Services Model Diffserv Can provide aggregate QoS Newer approaches Core-stateless schemes

Integrated Services Approach Guaranteed service Controlled load service Best effort service

Guaranteed Service Similar to a leased line Hard guarantees on bandwidth and maximum delay Addressed toward critical applications Advantages? Disadvantages?

Controlled load service Service provided equivalent to that of an “unloaded” network Admission control necessary No hard guarantees for bandwidth, delay, or loss! Advantages? Disadvantages?

Best-effort Service Currently supported in the Internet No guarantees whatsoever Advantages? Disadvantages?

Elements of QoS Flow specifications Routing Reservations Admission control Packet scheduling

Specifications and Buckets Leaky Bucket r Token Bucket r B

QoS Routing Determine a path through a network that can satisfy requirements Bandwidth a min-parameter => determining any one path simple Shortest widest vs. Widest shortest Delay an additive-parameter => cost based routing

RSVP: Resource Reservation Protocol Signaling protocol used to convey specifications of flow and desired quality of service Token bucket specification used in both Tspec (flow characteristics) and Rspec (desired QoS characteristics)

Admission Control & Scheduling Can a router accommodate the new incoming request without violating service to existing flows it serves? If yes, admit and appropriately configure scheduling algorithm If not, reject

Recap Quality of Service Integrated Services Guaranteed Controlled Best-effort Traffic shaping/policing with buckets

Integrated Services Disadvantages Per-flow state and processing at every router in the network Not scalable with increasing number of flows (in transit routers) and line-speeds

Differentiated Services Architecture Goal: To provide quality of service while ensuring scalability with increasing number of flows and line-speeds Approach: Have aggregate behaviors at the core with any per-flow state maintenance and processing done at the edges

Diffserv Approach Other autonomous domains Other autonomous domains Bandwidth Broker Admission control Edge Router: Policing, Shaping, & Marking Per-flow state and processing Core Router: Forwarding based on PHBs No Per-flow state and processing Other autonomous domains Other autonomous domains

Diffserv Service Classes Premium Service Assured Forwarding Service Best effort Service

Diffserv Classes - Premium Strict admission control Can still experience delays if a router has multiple incoming links Policing and Shaping done at the ingress points Similar to controlled-load service with no bursts

Diffserv Classes – Assured Forwarding Loose admission control More efficient Delays and drops possible during congestion Policing at the ingress points (no shaping) Non-conforming packets let through without marking

Intserv and Diffserv Intserv at the access and edge networks Diffserv at the core and transit networks RSVP can still inter-operate with the diffserv architecture When request arrives at ingress point, redirected to Bandwidth broker and forwarded to egress router

Diffserv Architecture Disadvantages No per-flow processing and hence no per-flow fine-grained QoS Example: no per-flow fairness possible in the diffserv architecture

Core-stateless QoS Goal: To provide per-flow fine grained QoS without maintaining any per-flow state at core-routers QoS Parameter: Rate fairness (delay fairness, bandwidth guarantees also possible) Approaches: CSFQ (Core-stateless Fair Queuing), Corelite

Core-stateless QoS Goal: To provide per-flow fine grained QoS without maintaining any per-flow state at core-routers QoS Parameter: Rate fairness (delay fairness, bandwidth guarantees also possible) Approaches: CSFQ (Core-stateless Fair Queuing), Corelite

Introduction Main Idea: Goals: - Achieve fair bandwidth allocations at the router without the implementation complexity usually associated with it. Goals: - Achieve fair allocation close to Fair Queueing and comparable or better than RED and FRED under most scenarios. - Reduce complexity by not having the core node maintain per flow state. - Approximate weighted FQ.

FIFO queueing with Drop Tail SERVER Disadvantages: Pushes congestion control out to end hosts (TCP) Introduces global synchronization when packets are dropped from several connections

Fair Queueing Disadvantage: Need to perform packet classification and maintain state and buffers on per-flow basis and perform operations on per-flow basis

Definitions Island of routers – a contiguous portion of the network with well defined interior and edges. Edge Router – computes per-flow rate estimates and labels the packets with these estimates. Core Router – uses FIFO queueing and keeps no per-flow state, employs a probabilistic dropping algorithm that uses the packet label and its own measurement of aggregate traffic. Stateless – absence of per-flow state at the core routers.

Island of Routers

CSFQ In an island of routers, edge routers compute per-flow rate estimates and label the packets with these estimates. Core routers use FIFO queueing and keep no per-flow state, they employ a probabilistic dropping algorithm based on packet labels and own aggregate traffic estimates. Bandwidth allocations using this method are approximately fair. Core routers keep no per-flow state and avoid using complicated packet scheduling and buffering algorithms, hence are easier to adopt.

CSFQ Assume that flow i has arrival rate ri(t) and the fair rate is a(t). If ri(t) < a(t), all of its traffic is forwarded. If ri(t) > a(t), then a fraction (ri(t) - a(t))/ ri(t) will be dropped; each packet of the flow is dropped with probability (1-a(t)/ri(t)). Thus the output rate of any flow i will be max(ri(t) ,a(t)).

CSFQ The problem now becomes how to calculate the flow rate ri(t) values and the fair rate a(t), without keeping per flow state in the core routers. Flow rates ri(t), are calculated at edge routers which keep per flow state and then insert the rate value inside the packet header of packets belonging to that flow.

CSFQ To estimate the fair rate a(t), an iterative procedure is used: core routers estimate aggregate arrival rate A and the aggregate rate of accepted traffic F (arrival rate – dropped packets). Based on these, the fair rate a is computed periodically as: - if there is no congestion (A<=C where C is the link’s capacity), then a is set to the maximum ri(t) - if the links are congested, then anew = aold*C/F

Source: Network Reading Group, Stoica CSFQ - Example Assume we have two flows f1 and f2, with rates r1 = 20 and r2 = 30 and the link’s capacity is C = 30. Initially let’s say that only r1 is active and the link is not congested, so a1 = 20. Then r2 becomes active. Since no packets were dropped, F = 50. Since A = 50>C, a2 = a1* C/F = 20 * 30/50 = 12 Therefore, for f1 (1-12/20 = 40%) of its packets are dropped while for f2 (1-12/30 = 60%) of its packets are dropped and F = 12+12 = 24 Since A>C, a3 = a2* C/F = 12 * 30/24 = 15 Now F = 30, and a4 = a3* C/F = 15 * 30/30 = 15. Therefore, a has converged to the right fair rate. Source: Network Reading Group, Stoica

Recap Intserv Diffserv Core-stateless Per-flow QoS Per-flow state/processing – not scalable Diffserv Coarse QoS No per-flow state/processing at all routers Core-stateless Scalable network model Per-flow QoS achieved

MPLS CoS (class of service) mechanism for the Internet Addresses speed, scalability, QoS (quality of service), and traffic engineering IETF defined framework that provides for efficient designation, routing, forwarding, and switching of traffic flows

MPLS – key functions mechanisms to manage traffic flows of various granularities – machines, flows, AS independent of layer 2 and layer 3 protocols means to map IP addresses to simple fixed length labels interfaces to existing protocols like RSVP and OSPF supports IP, ATM, and frame-relay

MPLS LSPs LSP – label switched path In MPLS, data transmission occurs on label switched paths LSP – a sequence of labels at each and every node along the path from the source to the destination LSPs established either before data transmission begins (control-driven) or upon detection of data (data-driven)

MPLS LSPs (contd.) Labels are distributed using LDP (label distribution protocol) Each data packet in an MPLS network carries a label High speed switching of data becomes possible as hardware can switch based on labels

Routers in MPLS LERs – label edge routers Operates at the edge of the network Responsible for establishing LSPs Adds/removes labels as traffic enters/leaves network LSRs – label switching routers High speed core router device Participates in establishing LSPs High speed switching of data using labels

Forward Equivalence Class (FEC) Class of packets that share same transport requirements All packets in a class treated the same Each LSR maintains a label information base (LIB) LIB comprises of FEC-to-label bindings

Labels A label identifies the path a packet should traverse within an MPLS network The label is typically carried in the layer 2 header or in a layer 2.5 header Label values have local significance only FR DLCIs (data link channel identifiers) or ATM VPIs/VCIs can be used as labels directly

Label Bindings Labels are bound to FECs based on some policy like destination unicast routing traffic engineering (RSVP-TE, CR-LDP) multicast virtual private network (VPN) QoS

Label Creation & Signaling [iec.org]

Puzzle Two twins A & B A always speaks the truth, and believes all true propositions (say 2+2=4) to be true, and all false propositions (say 2+2=3) to be false B always lies, and believes all true propositions to be false, and all false propositions to be true You meet one of the twins. How many questions do you need to identify which twin he is?