RD-QoS – The Integrated Provisioning of Resilience and QoS ......direct mapping of resilience attribute to MPLS recovery options •MPLS Traffic Engineering resource efficient resilience
Post on 11-Apr-2020
12 Views
Preview:
Transcript
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
1
RD-QoS – The Integrated Provisioning of Resilienceand QoS in MPLS-based Networks
Achim AutenriethMunich University of Technology
Institute of Communication Networks
Email: Autenrieth@ei.tum.de
IEEE International Conference on Communications (ICC)New York, USA
May 1, 2002
Andreas KirstädterSiemens AG, Corporate Technology
Information and Communication
Andreas.Kirstaedter@mchp.siemens.de
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
2
Outline
½ Introduction
½ Resilience Differentiated QoS (RD-QoS)
½ RSVP / DiffServ Resilience Signaling
½ Interworking with MPLS Recovery
½ Case Study and Results
½ Conclusion
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
3
Introduction
Resilience
Behavior under fault conditionsFault detection, failure notification,recovery and service restoration
MPLSsupports
QoSBehavior under normal conditions
Resource management, trafficmanagement (marking, shaping,queuing, metering)
MPLS offers various resilience optionsProtection Switching / Restoration, Local / Global Scope, …
Advantages of MPLS recovery are:Resource efficiency, recovery granularity, protection flexibility
QoSBehavior under normal conditions
Resource management, trafficmanagement (marking, shaping,queuing, metering)
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
4
Problem Definition
• MPLS recovery must be compared to optical network recovery
⇒ MPLS recovery should utilize its benefits to the most extent
• Moreover, service providers should be able to charge for higher
resilience as a value-added service
⇒ Services should be protected with the required level of resilience
But: How can this level be identified?
Resilience requirements (resilience attribute) should beincluded in the QoS signaling (like bandwidth and delay)
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
5
Resilience-Differentiated QoS
Extended quality-of-service definition: extend thestandard QoS-metrics (bandwidth, delay, delay jitter)
with resilience requirements of IP service classes
Extended quality-of-service definition: extend thestandard QoS-metrics (bandwidth, delay, delay jitter)
with resilience requirements of IP service classes
Resilience attribute• included in QoS signaling between application and network.
• depending on QoS architecture (IntServ, DiffServ) on a perflow or on a per packet basis.
• mapped to MPLS FECs with appropriate recovery options
4 Resilience Classes proposedmainly distinguished by recovery time requirements
RC1 - High RC2 - Medium RC3 - Low
10 - 100ms 100ms - 1s 1s - 10s pre-emption
RC4 - None
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
6
RD-QoS Network Model
FEC1
Core NetworkMPLS / DiffServ
Resilience mechanisms &Traffic Engineering
FEC2
Access networksDiffServ / RSVP
Resilience signaling &resource management
MPLS: MultiprotocolLabel SwitchingRSVP: ResourceReservation ProtocolDiffServ: DifferentiatedServicesFEC: ForwardEquivalence Class
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
7
RD-QoS Building Blocks
• Extended QoS architecture
resilience signaling between application and network
• QoS Resource Management and Traffic Conditioning
takes resilience attribute into account
• Recovery Mechanisms
provided by MPLS
• Interworking of RD-QoS with MPLS
direct mapping of resilience attribute to MPLS recovery options
• MPLS Traffic Engineering
resource efficient resilience provisioning
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
8
RD-QoS Signaling
DiffServ Network
Core routerTraffic SchedulingQueue Management
RM
Edge router (DS boundary)
Resource Management &Traffic ConditioningClassification, Policing,Marking, Shaping
with Resilience Attribute
In case of failure, non-resilient
packets can be treated
out-of-contract
DiffServ
QoS request with resilienceattribute is signaled throughnetwork-> Resource ManagementProtection: Signaling is doneon disjoint routes with explicitrouting
Network withRSVP-TE sign.
PathResv
RSVP-TE RMQoS request with resilienceattribute is signaled throughnetwork-> Resource ManagementProtection: Signaling is doneon disjoint routes with explicitrouting
Network withRSVP-TE sign.
PathResv
RSVP-TE RM
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
9
MPLS Recovery MechanismP
rote
ctio
nR
esto
rati
on
A
B C
DE F
GH I
Segment Prot. (Haskin)
A
B C
DE F
GH I
Fast Reroute
A
B C
DE F
G H I
Global Restoration
A
B C
DE F
GH I
Local Restoration
A
B C
DE F
G H I
Local to Egress Rest.
A
B C
DE F
GH I
Path Protection
Global Segment Local
Recovery Scope
Rec
ove
ry m
od
el
P1 P2 P3
R1 R2 R3
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
10
Resilience Class
RC1 RC2 RC3 RC4
Resilience requirements
High Medium Low None
Recovery time
10-100 ms 100ms - 1s 1s - 10s n.a.
Resilience scheme
Protection Restoration Rerouting Pre-emption
Recovery path setup
pre-establishedon-demand immediate
on-demand delayed
none
Resource allocation
pre-reserved on-demand (assured)
on-demand (if available)
none
QoS after recovery
equivalentmay be tempo-rarily reduced
may have reduced QoS
none
Interworking of RD-QoS with MPLS
Resilience classes are mapped to MPLS recovery optionsResilience classes are mapped to MPLS recovery options
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
11
RD-QoS Traffic Engineering
link
ban
dw
idth
RC1a
RC2a
RC3
RC4RC2b
RC1b
RC1: Protectiona: activeb: backup
RC2: Restorationa: activeb: backup
RC3: ReroutingRC4: Pre-emption
where:
• Offline MPLS Traffic Engineering with resiliencedifferentiation
• Used resources (guaranteed bandwidth)calculated on each link for the 4 resilience classes
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
12
RD-QoS Case Study
Network Scenario• Northern Italian research network• 16 nodes, 36 links • Demands between a pair of nodes
between 1 Gb/s and 16 Gb/s
4 Service Ratio Scenarios• 100% Best-effort traffic (RC3)• RD-QoS traffic with 10% RC1,
20% RC2, 40% RC3 and 30% RC4• 100% RC2 traffic (restoration)• 100% RC1 traffic (protection)
3 Protection and 3 Restoration mechanisms• P1: Path protection P2: Segment prot. P3: Link protection • R1: Global rest. R2: Local to egress rest. R3: Local rest.
GEN
SAV
ALETOR MIL
MIL2BRE VIC
VEN
BOL
FIRPIS
ROM
ANC
PIA VER
GEN
SAV
ALETOR MIL
MIL2BRE VIC
VEN
BOL
FIRPIS
ROM
ANC
PIA VER
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
13
0
200
400
600
800
1000
1200
1400
1600
1800
2000
RC1a RC2a RC3 RC4 RC1b RC2b
Case Study Results
A
Case Study Scenarios
Tot
al u
sed
reso
urce
s (
Gb/
s)
P1: Path protection R1: Global rest. P2: Segment prot. R2: Local to egress rest.P3: Link protection R3: Local rest.
B C D
P1
R1 R2 R3
E F G
P2
R1 R2 R3
H I J
P3
R1 R2 R3
K L M
R1
R2
R3
N O P
P1
P2
P3
Global recoverymechanisms have
better resourceefficiency than
local mechanisms
Global recoverymechanisms have
better resourceefficiency than
local mechanisms
RD-QoS has significantlybetter resource efficiency:
11%-33% recovery resourceratio compared to 115%-207% with full recovery
RD-QoS has significantlybetter resource efficiency:
11%-33% recovery resourceratio compared to 115%-207% with full recovery
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
14
Conclusionsu RD-QoS architecture extends QoS signaling with resilience
requirements of IP services to achieve flexible resilienceprovisioning
u 4 Resilience Classes proposed, primarily distinguished byrecovery time requirements
u RD-QoS achieves high resource efficiency for the cost ofincreased complexity (additional resilience attribute)
The current trend is clearly towards a service-driventransport architecture. The resilience requirements
should therefore be included in the QoS signaling likebandwidth and delay
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
15
Thank you for your attention.
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
16
Resilience Requirements of IP Services
• Resilience requirements of IP services are orthogonal totheir ”classical” quality-of-service requirements(bandwidth, delay, delay jitter)
highlow
highmission-critical VoIP and
multimedia servicesstandard VoIP and
multimedia services
low
database transactions, mission-critical control terminals, e-commerce
applications
e-mail, FTP, standard WWW
Resilience requirements
QoSRequire-ments
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
17
Resilience Classes
Proposed Resilience Classes RC1 - RC4:
RC1: High Resilience Requirements: 10 – 100ms recovery timeUse of MPLS protection switching or Fast Reroute
RC2: Medium Resilience Requirements: 100ms – 1s recovery timeMPLS Restoration with on-demand backup path establishment
RC3: Low Resilience Requirements: 1s – 10 s recovery timeNo resources are reserved / allocated in advance. Traffic recoveryrequires rerouting and resource reservation.
RC4: No Resilience Requirements: pre-emptionCorresponding to low-priority, pre-emptible traffic. Packets may bediscarded in case of failures.
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
18
Multiprotocol Label Switching (MPLS)
u MPLS integrates Layer 3 Routing with Layer 2 Switching
u Connection-oriented characteristic: hop-by-hop IP routingreplaced by label switching
u Packets are assigned to Forward Equivalence Classes (FEC) onlyonce at the network ingress
u Packets follow a pre-definedLabel Switched Path (LSP)
u Signaling protocols forpath setup:CR-LDP & RSVP-TE LSR LSR
LSRLSR
LSR
LER
LSR
LSRLER
FEC1
MPLS domain
FEC2
Assignment of different pathsfor flows with same source
and destination address
Assignment of different pathsfor flows with same source
and destination address
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
19
MPLS Recoveryu MPLS Recovery is currently a key research issue in the IETFu Several drafts are published which present recovery
mechanismsu “ Framework for MPLS-based Recovery ” defined in
[draft-ietf-mpls-recovery-frmwrk-03.txt]u Well known resilience concepts from SDH and ATM Recovery
are mapped to MPLS
Benefits from MPLS Recovery• Finer recovery granularity (compared to Layer 1 recovery)• Protection Selectivity based on Service Requirements
possible• Efficient and flexible resource usage (e.g., recovery path
may have reduced performance requirements)• Allows end-to-end protection of IP services
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
20
MPLS Recovery OptionsRecovery models
Resource Allocation
Resource Use
Path Setup
Recovery Scope
Recovery Trigger
Protection Switching Restoration (MPLS Rerouting)
Reserved-on-demandPre-reserved
Dedicated resources Shared resources Extra-traffic allowed
Pre-established Pre-Qualified Established-on-demand
Local RepairGlobal Repair Segment Repair
Automatic Input(internal signals)
External commands(OAM signaling)
Recovery models
Resource Allocation
Resource Use
Path Setup
Recovery Scope
Recovery Trigger
Protection Switching Restoration (MPLS Rerouting)
Reserved-on-demandPre-reserved
Dedicated resources Shared resources Extra-traffic allowed
Pre-established Pre-Qualified Established-on-demand
Local RepairGlobal Repair Segment Repair
Automatic Input(internal signals)
External commands(OAM signaling)
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
21
Path ProtectionProtection switching, pre-established, global scope, pre-reservedProtection switching, pre-established, global scope, pre-reserved
A E
GH
F
G
D
CB
p-LSP
w-LSP
+ Single backup LSP per working LSP– Failure signaling required+ Node failures covered
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
22
Fast Reroute [Haskin]
u Alternative recovery LSP set up from the last-hop LSR in reverse direction to the ingress
LSP and along a node-disjoint pathto the egress LSP
Source: [draft-haskin-mpls-fast-reroute-01.txt]
Protection switching, pre-established, pre-reserved,local switching, global recovery
Protection switching, pre-established, pre-reserved,local switching, global recovery
A E
GH
F
G
D
CB w-LSP
p-LSP
+ Single backup LSP per working LSP– No failure signaling required
+ Node failures covered– High spare capacity requirement
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
23
Link Protection
Protection switching, pre-established, local scope, pre-reservedProtection switching, pre-established, local scope, pre-reserved
A E
GH
F
G
D
CB
w-LSP
p-LSP3
p-LSP2
p-LSP1
– Multiple backup LSPs per working LSP+ No failure signaling required– Node failures not covered
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
24
Path Restoration
Restoration, established on-demand, reserved on-demand, global scopeRestoration, established on-demand, reserved on-demand, global scope
A E
GH
F
G
D
CB
w-LSP
– Failure signaling required+ Node failures covered
+ Alternative LSPs distributed over network => high spare capacity efficiency
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
25
Failure to Egress RestorationRestoration, pre-established, pre-reserved,
local switching, global recoveryRestoration, pre-established, pre-reserved,
local switching, global recovery
A E
GH
F
G
D
CB
w-LSP
+ No failure signaling required+ Node failures covered
o Between local and global routing => average spare capacity efficiency
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
26
Link Restoration
Restoration, established on-demand, reserved on-demand, local scopeRestoration, established on-demand, reserved on-demand, local scope
A E
GH
F
G
D
CB
w-LSP
+ No failure signaling required– Node failures difficult to cope with
– Alternative LSPs locally routed => lower spare capacity efficiency
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
27
RSVP-TE RC1 Protection Signaling5'�4R6
u Application signals resilience requirements to the network in addition to classical QoS requirementsu Network (additionally) reserves an alternative and disjoint route for the flow (e.g., with explicit routing)u Link or node failure: Traffic is sent over alternative route
MPLS domain
RSVP-TE signals LSP setup for RC1 through network1+1, 1:1 protection: Signaling is done on disjoint routes
PathResv
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
28
0,0
200,0
400,0
600,0
800,0
1000,0
1200,0
1400,0
1600,0
1800,0
2000,0
A B C D E F G H I J K L M N O P
RC1a RC2a RC3 RC4 RC1b RC2b
PANEL Case Study Results
100%
RC
3P1
/ R
1P1
/ R
2P1
/ R
3
P2 /
R1
P2 /
R2
P2 /
R3
P3 /
R1
P3 /
R2
P3 /
R3
R1
R2
R3
P1P2
P3
Simulation Scenarios
Tot
al u
sed
reso
urce
s (
Gb/
s)
P1: Path protection
P2: Haskin
P3: Link protection
R1: Global rest.
R2: Local to egress rest.
R3: Local rest.
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
29
COST Case Study Results
0
2000
4000
6000
8000
10000
12000
14000
16000
A B C D E F G H I J K L M N O P Q R S T U V
RC1a RC2a RC3 RC4 RC1b RC2b
100%
RC
3P1
/ R
1P1
/ R
2P1
/ R
3P2
/ R
1P2
/ R
2P2
/ R
3P3
/ R
1P3
/ R
2P3
/ R
3R
1R
2R
3P1 P2 P3
R1
R2
R3
P1P2
P3
Simulation Scenarios
Tot
al R
eser
ved
Net
wor
k C
apac
ity (
Gb/
s) P1: Path protection P2: Haskin P3: Link protection
R1: Global rest. R2: Local to egress rest. R3: Local rest.
Munich University of TechnologyInstitute of Communication NetworksProf.Dr.-Ing. Jörg Eberspächer
Achim AutenriethAutenrieth@ei.tum.deNK
L
30
Benefits
Interworking of RD-QoS with MPLS allows a direct mapping of RD-QoSclasses to MPLS LSPs with different protection levels according to the
negotiated resilience requirements
Interworking of RD-QoS with MPLS allows a direct mapping of RD-QoSclasses to MPLS LSPs with different protection levels according to the
negotiated resilience requirements
• RD-QoS as an integrated approach for the provisioning ofend-to-end QoS and Resilience
• Direct mapping of Resilience Classes to MPLS recovery optionspossible
• Applications define their resilience requirements⇒ protection flexibility⇒ efficient resource usage
• QoS requirements of high resilience traffic can be met in case ofnetwork failures
• RD-QoS as an integrated approach for the provisioning ofend-to-end QoS and Resilience
• Direct mapping of Resilience Classes to MPLS recovery optionspossible
• Applications define their resilience requirements⇒ protection flexibility⇒ efficient resource usage
• QoS requirements of high resilience traffic can be met in case ofnetwork failures
top related