320 likes | 417 Views
A Differentiated Services Architecture for the Internet. References - K. Nichols, V. Jacobson, L. Zhang - D. Clark, J. Wroclawski Presented by Liping Zhang. Overview. Introduction Two different service types: implementation and problems
E N D
A Differentiated Services Architecture for the Internet References - K. Nichols, V. Jacobson, L. Zhang - D. Clark, J. Wroclawski Presented by Liping Zhang
Overview • Introduction • Two different service types: implementation and problems • Two-bit differentiated services architecture • Problems with end-end bandwidth allocation based on level of marked traffic • Discussion
Introduction • Why do we need differentiated services? • Different users • Different applications • Service allocation • For example, one goal of assured service is to allocate the bandwidth of the Internet to different users in a controlled way during periods of congestion
How to describe a service • What is provided to the customer • E.g., 1 Mbps, continuously available • To where is this service provided • A single destination • A group • All nodes on local provider • Everywhere • Level of assurance provided to service • What level of performance uncertainty can user tolerate
Two distinct service types • Assured service • D. Clark • Premium service • V. Jacobson
Assured service • Provide different levels of best-effort service at times of network congestion • Expected capacity • “In” packets unlikely to be dropped • “Out” packets - no assurance • Queuing • Best effort
Mechanism for assured service Counter Counter Out- and in- dropper First-hop Host Marking packets according to the service profile RIO scheme, packets are treated preferentially
RIO algorithm • RED - Random Early Detection • Packets dropped with low but increasing probability as queue grows; instead of waiting until it is full and dropping all new packets • RIO • Run two RED algorithms for “in” and “out” with different dropping frequencies
Premium service • Fundamentally different Internet best effort service - high priority traffic has its own queue in routers • Shaped, hard-limited to provisioned peak rate • No bursts are injected into net • Virtual wire, available whenever needed • Regular flow pattern, no queuing • Shared, with best-effort
Mechanism for premium service Intra-network Host Router First-hop H-Q: premium, no dropping L-Q: best effort, dropping on congestion
Two-bit differentiated services architecture • Deploying both services • More bits available in IP header, why not both • Forwarding path mechanisms • Leaf routers • Input interface: a traffic profile • Output interface: two queues (HQ, LQ) • Intermediate routers • Only have forwarding function • Border routers • A Profile Meter at the input interface
Traffic flow from end-host to ISP Company A Internal Router Host 2 Border Router 1 First-hop Router 3 ISP Border Router
Forwarding path primitives • General classifier • In leaf routers, transport-level signature matching • Bit-pattern classifier • Performs a two-way decision based on bit-pattern • Bit setter • A- and P-bits must be set or cleared in several places • Priority queues • Shaping token bucket • At the leaf router for Premium traffic • Policing token bucket • At border router, for both P and A services
Block diagram of leaf router input functionality Marker 1 Flow1 Flow N Arriving packet Marker N Forwarding Engine Clear A&P bits Packet Classifier Best Effort
Markers to implement the two different services Wait for token Set P bit Packet Input N Test if token Set A bit Packet Input Y
Router output interface for two-bit architecture P-bit set? High-priority No If A-bit set? Inc a_cnt Low-priority If A-bit set? dec a_cnt RIO queue management
Border router input interface Profile Meters Token available ? N Clear A bit A set Y Forwarding Engine Is packet marked ? Not marked Y P set N Token Available Drop Packet
Passing configuration information • Request to the leaf router • Average rate, burst, service type (P or A) • Ways of passing the message • RSVP, SNMP, network administrator • Authenticating the sender
Architectural framework for marked traffic allocation • Preconfiguring of usage profiles is practical • Paying for level of service that is always available • Allocation follows organizational hierarchies • Each organization must be responsible for its DM • Only bilateral agreements work
Bandwidth Brokers (BB) • Roles • Allocating and controlling bandwidth shares • Responsibilities • Parcel out a region’s marked traffic allocation and set up the leaf routers within the local domain • Managing messages sent across boundaries to adjacent region BBs
Examples • A statically configured example with no BB message exchanged • A statically configured example with BB messages exchanged • Dynamic allocation and additional mechanism
RSVP and BBs • Existing bilateral relations between BBs of adjacent trust regions are necessary for resource allocation • A few bits in the packet header are used to mark the service class • RSVP resource setup: hop-by-hop • Use RSVP between two adjacent ISPs (BB1/BR1 and BB2/BR2)
Discussion • Extensibility of the current 2-bit architecture • Service allocation for multicast • Who should request the service • Sender or receiver • Deployment issues • Security issues
2-bit differentiated services architecture • Providing Controlled-Load and Guaranteed service • P service for C-L service • A constrained case of C-L service • P service for G service • The service model of P service fits G service model