Transcript IntServ and DiffServ

IntServ and DiffServ
School of Electronics and Information
Kyung Hee University.
Choong Seon HONG
<[email protected]>
Kyung Hee University
Quality of Service (QoS)
A major driving force in Internet evolution
Not simply defined - means many things to
many people
Has sense of predictable network behaviour
Central idea is provision of network
resources that an application requires to
perform adequately
Kyung Hee University
Quality of Service (QoS)
 What is Quality-of-Service?
• Quality of service (QoS) is a concept by which
applications may indicate and even negotiate their
specific service requirements to the network
 Why is this an issue?
• The default service in many packet networks is to
give all applications the same service, and not
consider any service requirements to the network.
This is called a best-effort service.
Kyung Hee University
Quality of Service (QoS)
 Who needs Quality-of-Service?
– Video and audio conferencing  bounded delay and
loss rate
– Video and audio streaming  bounded packet loss rate
– Time-critical applications (real-time control) 
bounded delays
– “valuable applications”  better service than less
valuable applications
Kyung Hee University
Quality of Service (QoS)
 How are Quality-of-Service requirements
specified?
• QoS parameters are
– Delay
– Delay Variation (Jitter)
– Throughput
– Error Rate
Kyung Hee University
Quality of Service (QoS)
 What is the granularity of QoS?
– Per-flow QoS
 Guarantees are specified and enforced for single
end-to-end data flow
– Aggregate QoS
Guarantees are specified and enforced for groups of
flows
Kyung Hee University
Types of QoS guarantees
 Deterministic QoS
– Service guarantees are enforced for all traffic
 For example, deterministic delay guarantees have the form:
Delay of a packet from flow X ≤ D
(D is called a delay bound)
 Statistical QoS
 Allows a certain fraction of traffic to violate the service
guarantees
 Prob [Delay of a packet from flow X ≤ D ] ≥1 - ε
Where e is a small number (e.g., ε = 10-6)ε
Kyung Hee University
 Classification and Scheduling
 Routers need to be able to
1. classify arriving packets according to their QoS
requirements
 Packet Classification
2. isolate traffic flows and provide requested QoS
 Packet Scheduling
Kyung Hee University
QoS is Generating a Confusing Array of Acronyms
QoS
CoS
Intserv
RSVP
Diffserv
MPLS
GMPLS
Kyung Hee University
Why Do We Need Such a Revolutionary Change?
 Current ‘best effort’ technology is essentially a quarter
of a century old
 Two factors driving the development of a new
generation of multimedia applications
commercialisation of the Internet
Increasing availability and decreasing cost of bandwidth
 No evidence of ‘free bandwidth’ scenario emerging
rejected in RFC1633 (1994) - still true
demand always rises to meet supply
Kyung Hee University
QoS is Not New
 Telephone network has QoS
economics and technology based on a single application
highly developed engineering
but one size fits all
 BISDN - an attempt by telephony world to generalise
network to encompass diverse applications
 ATM technology - first full exploration of QoS on
demand concepts
Kyung Hee University
Quality of Service and Resource Management
 Fundamental resource is output link rate
 Access managed via scheduling discipline
 Bursty input traffic held in buffers
adds delay and jitter
overflow causes packet loss
 These factors determine QoS at network level
 Optimise via buffer management and scheduler
parameter setting
Kyung Hee University
QoS in the Internet
 Internet Engineering Task Force (IETF) is
evolving QoS support mechanisms for the Internet
- two approaches
The Integrated Services Internet
• QoS for individual microflows
• perhaps too complex for large networks - won’t scale easily
Differentiated Services - more scaleable
• lose sight of individual microflows - Behaviour Aggregates
Kyung Hee University
 Integrated Services (Intserv)
 QoS approached via end to end services
best effort - current performance standard
controlled load - lightly loaded network performance ‘soft’ delay bound
Guaranteed - ‘hard’ bandwidth and delay bounds
 Traffic conformance to agreed form expected
‘token bucket’ model - policing if nonconforming
 Resources reserved in routers - RSVP
more complex set of functions than ATM
Kyung Hee University
RSVP is Dead!
 Earlier reports of RSVP’s death were somewhat
exaggerated
 Nevertheless there is a major problem with
Intserv- fatal in the eyes of some
 Management of router resources requires each
router to maintain per flow ‘state’
 Creates ‘state explosion’ in the interior routers of
core networks - perhaps confine to edges
Kyung Hee University
 Differentiated Services (Diffserv)
 Driving philosophy of the Internet has been to
minimise complexity in the core network - push
complexity and intelligence to the edge nodes.
 Differentiated Services concept strives to maintain
this philosophy while recognising the need to
provide some levels of Quality of Service
 First widely deployed QoS mechanism
Kyung Hee University
Differentiated Services
Kyung Hee University
Content
 Intserv/RSVP
 Differentiated Service Paradigm
 Per-Hop Behavior & Codepoint
 Premium Service
 Assured Forwarding PHB Group
 Resource Manager : Bandwidth Broker(BB)
 Boundary Mechanisms
 Diffserv WG
Kyung Hee University
Internet Integrated Service Model
Guaranteed Quality of Service
 Motivation
 for applications intolerant of late data
 hard real time requirements
 End-to-End behavior
 an assured level of bandwidth
 a delay-bounded service with no queueing loss
 firm maximum on end-to-end delay
 not control the minimal or average delay
 no jitter control
Kyung Hee University
 Internet Integrated Service Model
 In order for a router to invoke Guaranteed Service for a
specific data flow it needs to be informed of the traffic
characteristics of the flow, Tspec, along with the
reservation characteristics, Rspec
 Tspec parameters
•
•
•
•
•
p ; peak rate of flow (bytes/second)
b ; bucket depth (bytes)
r ; token bucket rate (byes/second)
m ; minimum policed unit (bytes)
M ; maximum datagram size (bytes)
 Rspec parameters
• R ; bandwidth, i.e. service rate (bytes/second)
• S ; Slack Term (ms)
Kyung Hee University
Internet Integrated Service Model
Controlled - Load Service
 Motivation
 for adaptive real-time applications (today’s internet)
 work well on unloaded nets but degrade quickly under overload conditions
--> mimics unloaded nets
 If the flow is accepted for Controlled-Load Service then the router makes
a commitment to offer the flow a service equivalent to that seen by a besteffort flow on a lightly loaded network
Kyung Hee University
Internet Integrated Service Model
Controlled - Load Service (cont’d)
 End-to-End behavior
 Tightly approximates the behavior visible to applications receiving besteffort service under unloaded conditions
 A very high percentage of packets delivered successfully
 Controlled Load has some fairly simple implementations, in terms of the
queuing systems in routers
 It is not suited to applications that require very low latency (e.g.
distributed VR systems and so forth).
Kyung Hee University
RSVP
RSVP Design Principles
 Receiver-Initiated Reservation
 a receiver
• choose the level of reservation
• initiate/keep reservation
 more flexible and scaleable than source-initiated reservation
• heterogeneous receivers
• dynamic membership change
 Separating reservation from packet filtering
 reservation
• amount of resources reserved for an entity
 packet filtering
• dynamically select packets that can use the resources
Kyung Hee University
 RSVP
Design Principles (cont’d)
 Maintain “Soft-state”
 dynamic status change (membership change)
 soft-state in switches and maintained by end users
 state in switches
• path state -- periodic path message from the source
• reservation state -- periodic reserv. msg from the receivers
 timeout driven deletion
• Reservations timeout if not refreshed periodically
 adaptability and robustness
 Protocol overhead
 reduce refreshing frequency
 merging path/reservation messages
Kyung Hee University
RSVP
Comparison of RSVP and ATM signaling
RSVP
ATM
Receiver generates reservation Sender generates connection request
Soft state ( refresh / timeout )
Hard state ( explicit delete )
Separate from route
establishment
Concurrent with route establishment
QoS can change dynamically
QoS is static for life of connection
Receiver heterogeneity
Uniform QoS to all receivers
Kyung Hee University
RSVP Message Types
1. Path_msg
S1
2. forwarding Path_msg
R1
R2
D1
S2
R3
3-1.Resv_msg
4. forwarding Resv_msg
D2
3-2. Resv_msg
Kyung Hee University
Internet Integrated Service Model
Integrated Services over Specific Link
Layers(ISSLL WG)
•
RSVP designed to work with any protocol
- Protocol must provide QoS support
- Examples: ATM, IP with Integrated Services
• IP integrated services with RSVP over ATM
– VC management ( traffic flow-VC)
• Data VC, RSVP signaling VC
– QoS translation
• mapping a QoS from the IIS model to a proper ATM QoS
• IIS over POTS
• IIS over LAN
Kyung Hee University
Intserv / RSVP QoS Approach
Service
Parameters
- Token bucket rule
- Guaranteed Service - Token bucket size
- Control Load Service - Packet rate
- Best-Effort Service
- Minimal policed unit
- Maximum packet size
Traffic
Management
- RSVP
- Admission Control
- QoS Routing
- Control of Traffic
Parameters
 Scalability problem
• Have to maintain forwarding state between receiver and transmitter
Kyung Hee University
Integrated Services Model
 Flow specification
 Routing
 Admission control
 Policy control
 Resource reservation
 Packet scheduling
Kyung Hee University
RSVP Functional Diagram
Host
Router
RSVPD
RSVPD
Routing
Process
Application
D
A
T
A
Packet
Classifier
Policy
Control
Policy
Control
Admissions
Control
Admissions
Control
Packet
Scheduler
DATA
Packet
Classifier
Kyung Hee University
Packet
Scheduler
DATA
What is a flow?
 Equivalent packets by some classification
RSVP: Set of packets traversing a network element that
are all covered by the same QoS request
 Packet classifier determines which packets belong
to which flows
IPv6 includes a flow label to ease classification
 ISP usage (UUNET)
Microflow: TCP or similar bandwidth connection
Macroflow: Large aggregates of packets flowing
between two superhubs
Kyung Hee University
Describing and Identifying a Flow
 Flowspec defines traffic parameters
Traffic parameters: bandwidth, buffering requirements
Uses token bucket specification
 Filterspec identifies packets in flow
Simplest filter: Source, Dest address/port pair
Data filter: classifies packets according to contents
Kyung Hee University
Resource Reservation
 Senders advertise using PATH message
 Receivers reserve using RESV message
Flowspec + filterspec + policy data
Travels upstream in reverse direction of Path message
 Merging of reservations
 Sender/receiver notified of changes
Kyung Hee University
RSVP UDP Reservation (1)
Kyung Hee University
RSVP UDP Reservation (2)
Kyung Hee University
Client Traffic Shaping
 Issue: Need traffic shaping to meet allocated
resources
 Source promises that data traffic will conform to a
particular shape
 Why describe and shape traffic?
Network knows what to expect, can manage traffic better
Better admission control decisions
Network can police flows
 Bursty traffic is costly to router, network
Kyung Hee University
Traffic Shaping Example
Data Queue
Flow 1
Flow 2
Data Queue
Kyung Hee University
Traffic Shapers
 Simple leaky bucket
Isosynchronous flow: regular intervals between packets
 Token bucket
Bursty flow
Kyung Hee University
Simple Leaky Bucket
Data
b
b = bucket size
r = rate data is sent onto network
r
 Sends data at fixed intervals onto network
 Bursts bigger than b are discarded
 Traffic never injected faster than r
 Can be used with cells or datagrams
Kyung Hee University
Token Bucket
r
b
b = bucket size in tokens
r = rate tokens are added to bucket
Data Queue
Data
 Sends bursty traffic onto network
 Bucket filled with tokens at rate r
 Data transmitted when enough tokens exist
 Allows bursts, but enforces upper bound
Kyung Hee University
Restrictions on Reservations
 Admissions
Is bandwidth available?
 Policy
Service guarantees give preferential access to network
bandwidth
Permissions
Pricing issues
 What are the policies of nodes on the path?
Policy data represents a scaling and security issue
Kyung Hee University
Resource Reservation Model
 Senders advertise using flowspecs
 RSVP daemons forward advertisements to
receivers, update available bandwidth, minimum
delay
 Receivers reservations use flowspec, filterspec
combination (flow descriptor)
 Sender/receiver notified of changes
 Reservations are merged in multicast case
Kyung Hee University
Reservation Styles
 Wildcard Filter (WF)
Shared reservation, select all upstream senders
Traffic from upstream senders shares a single pipe
Appropriate for audio
 Shared Explicit (SE)
Shared reservation, explicit sender selection
Appropriate for audio
 Fixed Filter (FF)
District reservations, explicit sender selection
Appropriate for video
Kyung Hee University
RSVP Flowspecs
Sender TSpec, Controlled Load Flowspec
...
Token Bucket Rate [r]
Token Bucket Size [b]
Peak Data Rate [p]
Minimum Policed Unit [m]
Maximum Policed Unit [M]
Guaranteed Flowspec
...
Token Bucket Rate [r]
Token Bucket Size [b]
Peak Data Rate [p]
Minimum Policed Unit [m]
Maximum Policed Unit [M]
Rate [R]
Slack Term [S]
Kyung Hee University
Packet Scheduling
 Implemented in hosts/routers to control link
allocation
 Queuing algorithms
Weighted Fair Queuing (WFQ)
Class Based Queuing (CBQ)
 Queue management
Random Early Detection (RED)
Kyung Hee University
Packet Scheduling
 Fair Queueing
 Attempts to implement a scheduler that serves all flows
with a backlog at the same rate
 Emulates a bitwise Round Robin scheduling algorithm
 Not completely trivial to implement Fair Queuing in a packet
network
Kyung Hee University
Weighted Fair Queuing (WFQ)
 Traffic placed into queues according to flow
specification, flow filter
 Fair queuing
Implements fairness of bit by bit scheduling on a per
packet basis
Gives queues a fair share of total bandwidth
 Weighted
Queue are not weighted evenly for scheduling
 Proven: adequate for Guaranteed Service
Kyung Hee University
Class Based Queuing (CBQ)
 Combines scheduling and link sharing
 Hierarchical link sharing
Hierarchical queues
Enables protocol, organization isolation
 Scheduling
Does not define a particular scheduling algorithm
General scheduler for low latency when no congestion
Link-sharing policing scheduler when congested
Scheduling per hierarchy
Kyung Hee University
CBQ Example
LINK
60%
40%
Company A
Company B
30%
RealTime
HTTP
FTP
telnet
IP
DECnet
20%
10%
20%
20%
Video
Audio
20%
10%
Kyung Hee University
Random Early Detection (RED)
 Random Early Detection (RED)
Queue management algorithm for congestion control
Random packet drops as average queue length increases
Can use Explicit Congestion Notification bit instead of
dropping packet
Works well for TCP
Useful for congested Controlled Load service
Kyung Hee University
Reservation Merging
(3) 50Kbs (7) 100 Kbs
R1
Reservations merge
as they travel up tree.
(6) 100 Kbs
R3
(2) 50Kbs
(9) 60Kbs
R4
(1) 50Kbs
Receiver
#1
R6
(8) 60Kbs
Receiver
#2
Kyung Hee University
(5) 100 Kbs
R7
(4) 100 Kbs
Receiver
#3
TSpecs, AdSpecs, and RSpecs
 Traffic source sends TSpec (Traffic Specification)
Consists of FlowSpec and AdSpec
 AdSpec updated to reflect network capabilities
Routers update minimum delay and maximum
bandwidth
Termed One Pass With Advertisement (OPSA)
 RSpec
Receiver uses Controlled Load or Guaranteed FlowSpec
to reserve network resources
Kyung Hee University
Problems with Merging Reservations
 Issue: who pays for service, how much?
 Merging different types of flows
Flow 1: Low delay, low bandwidth
Flow 2: High delay, high bandwidth
Flow with low delay, high bandwidth satisfies Flows 1
and 2, but it may cost much more than Flow 1 or 2.
 Only certain flows can be easily merged given
price constraints
Kyung Hee University
Reservation Merging and Price
Merged Reservation:
High Bandwidth,
Low Latency
Reservation 2:
High Bandwidth,
High Latency
Reservation 1:
Low Bandwidth,
Low Latency
Price: Darker = More Costly
Latency
Kyung Hee University
RSVP Routing Problems
 Routing is separated from admission control
 If route changes, reservation must be made along
new route
New reservation takes time to setup
New reservation might fail
Old route could still be working fine
 Route pinning
Always use the route where reservation is in place
Kyung Hee University
Routing Problems (cont’d)
 Reservation failure
Primary route has inadequate bandwidth although
secondary has enough
 Telephone system has a crankback feature
Allows secondary routes to be considered if reservation
on primary route fails
 ATM
Routing combined with admission control
Kyung Hee University
Usage and Implementation
 RSVP is not widely available
Best effort delivery across links with no RSVP services
Reservation flag to specify that traffic traveled over a
non-RSVP link
 Some links will have guaranteed performance for
some traffic, but not all
Policy issues at boundaries of networks
Kyung Hee University
Differentiated Service Paradigm
 Complicated operation moves to edge, and stateless in
network interior
“ push all the state to the edges, and force all perconversation work (e.g., shaping, policing) to the edges”
 Setting a specific part in an edge node and administrative
boundaries
DS(differentiated service) field
 How to forward according to a specific field of input
packet
Per-Hop Behavior
 According to service rule that is previously promised
Traffic Conditioning
Kyung Hee University
Traffic Conditioning
 Traffic conditioning mechanisms at the network
boundary need to enforce that traffic from a flow
adheres to its specification
 Policing
Drop traffic that violates the specification
 Shaping
Buffer traffic at network entrance that violates
specification
 Marking
Mark packets with a lower priority or as best effort, if
the traffic specification is violated
Kyung Hee University
Traffic Conditioning
 The most popular traffic conditioning algorithm is the leaky
bucket
Kyung Hee University
Per-Hop Behavior & Codepoint
IPv4 Header (first 32bits)
4-bit
version
4-bit
header
length
8-bit type of
service (TOS)
6-bit DSCP for
Per-Hop Behavior
16-bit total length (in byte)
2-bit
CU
Currently
Unused
-- DS field in IPv4--
Kyung Hee University
Per-Hop Behavior & Codepoint
Default PHB
 Current best effort forwarding
 codepoint : 000000
Class Selector PHB
 for backward compatibility (IP precedence field)
 codepoint : xxx000
 relative service quality
Kyung Hee University
Premium Service
 Providing resources according to Peak capacity
Profile
Static allocations on peak rate with no statistical
sharing
Small percentage of the total network capacity allocate
for Premium service
Much higher cost (First class in aircraft)
 Commercial applications for Premium service
Video broadcasts, voice-over-IP, VPNs, etc.
Kyung Hee University
Premium Service
Company A
Internal
Router
Premium packet flow
restricted to r rate
per sec
Host
First-Hop
Router
Unmarked
packet flow
Border
Router
Packet in premium
flows have bit set
ISP
Border
Router
-- Premium traffic flow from end-host to organization’s
ISP -Kyung Hee University
Premium Service
 Forwarding Path Primitives
General Classifier
• A transport-level signature matching based on a tuple in the
packet header
Bit-pattern Classifier
• A simple two-way decision based on whether a particular bitpattern in the IP header is set or not
– Ex) ‘P’ bit
Bit setter
• Sets the appropriate bits of the IP header to a configured bitpattern would be the most general
Kyung Hee University
Premium Service
 Priority queues
(At least) Two levels of simple priority queuing
• The high priority queue for Premium traffic
 Shaping token bucket
Forward an arriving packet if there is a token present in
the bucket, otherwise the packet is enqueued until the
bucket contains tokens sufficient to send it
• Used in Leaf router
 Policing token bucket
Never hold arriving packets, but check token
availability
• Used in Border router
Kyung Hee University
Assured Forwarding PHB Group
 N AF classes
 M drop precedence level
 at this point
4 classes, 3 drop precedence in each class
 Example Service : Olympic service
Kyung Hee University
Olympic Service
Gold Service Class
Silver Service Class
Bronze Service Class
 Packets assigned to Gold service class experience lighter load than
packets assigned to the silver class
 Packets within each class may be further separated by the drop
precedence
 Drop precedence level by using a dual leaky bucket traffic policer :
committed burst, excess burst
Kyung Hee University
Resource manager
: Bandwidth Broker (BB)
 A logical entity residing in each administrative
domain
Managing internal demands & resources according to
the policy database (who can do what when)
setting up & maintaining bilateral agreement with
neighbor domains
• bookkeeping how much traffic entering which border router &
going out which border router
 Today’s BB : network administrators & operators
would like to automate over time
Kyung Hee University
Bandwidth Broker (BB)
 Dynamic bandwidth allocation and TCA
management
Kyung Hee University
Choices for implementation
 Adequate provisioning
 Manual configuration
not that different from static routing
 Using some setup protocols
inter-domain : BB-to-BB
intra-domain : RSVP as a ready candidate
Kyung Hee University
The TCA
 One per customer - two parts
 Constraint TCA
protects the provider’s resources
DS field : metering profile : disposition of n/c traffic
quantitative service levels will also include destination
address
 Fine grain TCA
specifies the fine grain traffic conditioning requested by
the customer
mf class. Criteria : mark : shaping profile : disposition
of n/c traffic
Kyung Hee University
Constraints
Resources implied by fine grain TCA are
constrained by those permitted in the
constraint TCA
Sum of shaping profiles for each mark must be
less than metering profile for corresponding DS
Kyung Hee University
Inter-domain DS
 A Service Level Agreement (SLA) includes a Traffic
Conditioning Agreement (TCA)
 Simplest way: an administrative issue
Kyung Hee University
Configuring Routers
 TCA (Traffic Conditioning Agreement)
constraint
fine grain
 PHB information
 Miscellaneous
interface configuration
routing configuration
etc.
Kyung Hee University
Configuring the Constraint TCA
 Specifies agreement between provider and
customer
 Relatively static
Configuration via SNMP, CLI, COPS, etc.
 Dynamic
Configuration via COPS or alternate ‘BB’ protocol
Could be triggered by high/low water marks
Kyung Hee University
QoS Policy Control
Kyung Hee University
Configuring the fine Grain TCA
 Ability to make frequent changes desirable
 No need to negotiate with provider
 Potentially error prone process
 Therefore - use a signaling protocol to do this
whenever possible : RSVP
 It will still be necessary to provision certain fine
grain entries
for these, use COPS, SNMP or CLI
Kyung Hee University
Signaled vs. Provisioned
TCA Entries
 Quantitative QoS apps use quantitative services
TCA entries configured by RSVP signaling or
provisioned
TCA entries specify egress points
 Qualitative QoS apps use qualitative services
provisioned only
 Should use separate marks (DS-field)
Kyung Hee University
Admission Control
 a/c necessary to prevent over-subscription
sum of qualitative entry shaping profiles leq than
metering profile for qualitative DS fields
• enforced at provisioning time(SNMP, CLI or COPS)
sum of quantitative entry shaping profiles leq than
metering profile for quantitative DS fields
• enforced by rejection or acceptance of RSVP signaling
messages
• if quantitative resources are provisioned, a/c must be enforced
both at provisioning and signaling times
• difficult to manage
Kyung Hee University
Diffserv WG status
 Closed
 RFC 2474, Definition of the Differentiated
Services Field (DS Field) in the IPv4 and IPv6
Headers
 RFC 2475, An Architecture for Differentiated
Service
 ---
Kyung Hee University
Diffserv / MPLS
 Diffserv
BA (Behavior Aggregate)
• BAS (BA Selector) ; forwarding queue behavior
• BM (Behavior Modifier) ; Dropping behavior
PHP
 MPLS
Single LSP
• Same BAS
• BM Carried as part of label encapsulation header
Kyung Hee University
Diffserv / RSVP
Policy Server
PDP : Policy Decision Point
PDP
PEP
PEP
PEP : Policy Enforcement Point
Diffserv Network
PEP
PEP
PEP
RSVP
RSVP
Diffserv / PHP / Codepoint
Kyung Hee University
PEP