Converging Protection and Restoration Strategies of the IP and Optical Layers to Support the Surviva

1
Converging Protection and Restoration Strategies
of the IP and Optical Layers to Support the
Survival of IP Services

• MPLS Next Generation Networking
• January 2001
• Dr. John Strand
• jls_at_research.att.com
• ATT Labs

2
Outline

• Restoration 101
• Objectives
• Restoration Requirements
• Effect Of Topology
• SONET/Optical layer protection/restoration
• Automatic Protection Switching (APS)
• Self-Healing Rings
• Mesh-based restoration
• Ring vs mesh
• IP layer restoration
• IP rerouting
• Protection/restoration using MPLS
• Convergence generalized MPLS for Optical Layer
  control plane
• A New joint IP/Optical Restoration Mechanism

3
Service Restoration ObjectivesTypical Commercial
Networks
1000
100
10
Objectives (secs)
Restoration Time
1
Ring (BLSR)
0.1
0.01
Service Hits
Standard Voice
Leased Lines
Frame Relay
IP Services

• New trends in IP services supporting real-time
  application, e.g. voice and video, mission
  critical data
• gt Require much faster restoration than
  traditional IP rerouting

4
Restoration 101
C
D
A
B
X
Y

• Restoration Requires
• Fault Detection
• Spare Inter-Office Capacity
• Switch Fabric To Put Failed Facility On This
  Capacity
• Control Logic To Reroute Failed Circuits

5
Effect Of Network Topology
Restoration Overbuild (Protection
Capacity/Service Capacity)
100
50
1/(N-1)
6
Effect Of Network Topology
Restoration Overbuild (Protection
Capacity/Service Capacity)
"Ring"
100
50
"Mesh"
1/(N-1)
7
Self-Healing Ring No Failures
S
C
B
S
P
S
P
P
A
D
P
P
S
S
P
F
E
S
Original Circuit
SONET Line Switched Ring Example
8
Self-Healing Ring ADM - No Failures
STS-1 Fabric
SERVICE WEST
SERVICE EAST
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
PROTECTION EAST
PROTECTION WEST
SONET Line Switched Ring Example
9
Self-Healing Ring Automatic Protection Switching
S
C
B
S
P
S
P
P
A
D
P
P
S
S
P
F
E
S
Original Circuit
SONET Line Switched Ring Example
Protection Switch
10
Self-Healing Ring ADM - Automatic Protection
Switching
STS-1 Fabric (60 ms switch time)
SERVICE WEST
SERVICE EAST
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
PROTECTION EAST
PROTECTION WEST
SONET Line Switched Ring Example
11
Self-Healing Ring Ring Switch
X
S
Original Circuit
SONET Line Switched Ring Example
Ring Switch
12
Self-Healing Ring ADM - Automatic Protection
Switching
STS-1 Fabric (60 ms switch time)
SERVICE WEST
SERVICE EAST
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
OC-48 Terminating Equipment
OC-48 Terminating Equipment
OC-48
OC-48
PROTECTION EAST
PROTECTION WEST
SONET Line Switched Ring Example
13
Optical Layer Restoration Architecture
Alternatives

• Optical APS - mimic SONET APS
• Optical Self-Healing Rings - mimic SONET rings
• Could be 2/4 wire, uni or bi-directional,
  path-switched or line-switched
• OLXC mesh-based methods
• centralized
• distributed
• quasi-centralized

14
Optical Layer SwitchingFree-Space Micromachined
Optical Switch
Switch Time lt 1 ms 8x8 is 1 cm x 1
cm Opportunities To Extend To Significantly
Larger Arrays On A Single Substrate Measured
Switching Times Under 1 ms (500ms)
Output fibers
Free-rotating switch-mirror array
Micro lens
Silicon substrate
Si substrate
Input fibers
15
Optical Layer SwitchingAn 8x8 Switch
Chip size 1 cm x 1 cm
But Fault Detection Localization Is An Issue
Source L-Y. Lin
16
Optical Layer Restoration Example Failure Detected
OC-48 connection (normal path)
LOS received by add/drop restoration process
begins
LOS received by add/drop restoration process
begins
Router
signaling channel
Router
X
WDM
OXC
OXC
OXC
X
WDM
B
add/drop port
C
A
working channel
restoration channels (uncommitted)
OXC
OXC
D
E
node span disjoint restoration path
(pre-calculated)
Reference R. Doverspike J. Strand, "Robust
Restoration In Optical Cross-Connects", INFORMS
2000
17
Optical Layer Restoration Example Restoration
Process
OC-48 connection (normal path)
Router
signaling channel
Router
X
WDM
OXC
OXC
OXC

X
WDM
B
add/drop port
C
A
working channel
restoration channels (uncommitted)
channel selection protocol one end of each link
controls selection of restoration channel
inventory
OXC
OXC

D
E
node span disjoint restoration path
(pre-calculated)
18
Optical Layer Restoration Example Restoration
Process
OC-48 connection (normal path)
Router
signaling channel
Router
X
WDM
OLXC
X
WDM
B
add/drop port
C
A
working channel
restoration channels (uncommitted)

D
E

• for 2-way failure, messages meet in middle
• protocol resolves channel selection
• connection restored when xconn made

node span disjoint restoration path
(pre-calculated)
19
IP Layer Rerouting
Chicago
San Francisco
New York
Atlanta
Dallas

• IGP LSAs update topology change
• Routing algorithm re-converges
• Take secs to mins
• Need built-in spare capacity,
• otherwise congestion occurs

20
Enhance IP Survivability by using MPLS

• Advantages
• Faster restoration than traditional IP rerouting,
  potentially in subsec
• Cost effect restoration by providing the required
  level of restoration for different classes of
  traffic
• Disadvantages
• Requires additional build-in spare capacity at
  layer 3, otherwise congestion could occur
• gt Larger/additional pipes and interfaces on
  routers

21
Effect Of Layering
IP on WDM
Current stack
TCP
TCP
Re-Transmittal
Routing Re-Convergence
IP
IP
VP/VC Restoration
Mesh Restoration
SDH/SONET Ring Restoration
Optical Restoration
WDM
WDM
22
Counter-Productive Protection Behavior
Link in Traffic
Routing table Revision (no link)
Routing table Revision (with link)
Link Rediscovered
ALARM
Link recovered through optical protection
Link Down
10s ms
10s seconds
10s seconds

• Instant response to Level 1 alarms in routers
  causes unnecessary routing activity, routing
  instability, and traffic congestion

Source RHK
23
Optical Layer Restoration MotivationReducing
Switching Costs
Relative Cross-Connect Cost
Location Of Cross-Connect Function
Router
Router 7.4
ADM (SDH RING) 4.7
ADM
Thru Wavelength 1.0
ADM
Thru OTS (No XC) 0
Optical XC
O T S
O T S
Thru OTS
However . . . The OXC Must Switch Large
Bundles Inefficient If Only A Small
Proportion Of Traffic Really Needs Restoration

Trade-Off Granularity vs Switch Cost/Gb Restored
24
Restoration Economics
Total cost
OC48 miles
R Proportion Of Connections Restored
References. R. Doverspike et al, "Transport
Network Architectures In An IP World", Infocom
2000.
25
Network Restoration Layer Characteristics
Network Factors
Optical Layer
SONET/SDH

IP/MPLS
Restoration Alternatives
APS, SHR, mesh restoration
APS, SHR, DCS-based restoration
Router-based rerouting
Unit Cost (/bit)
Lowest
Medium

Highest
Granularity
Coarsest (gt 1 Gb/sec)
Medium
Finest (LSP)

Restoration Speed
Few 100 ms
100 ms

Few seconds - minutes
Standards
ITU, ANSI/T1
ITU, ANSI/T1
IETF
Technology Maturity
Some available
Available

Available
26
Network Survivability

• Definition an aspect of network reliability that
  quantifies the performance of a network under
  failure conditions

Fig-1 Network Survivability Planning
27
Generalized MPLS for Optical Layer Control Plane

• Utilize the common suite of protocols for setting
  up, maintaining, and restoring lightpaths in
  Optical Layer
• Provide extensions to address Optical Layer
  unique features and requirements
• Provide potential to integrate protection/restorat
  ion in IP and Optical Layers
• gt Ongoing work in IETF, OIF, ITU/ANSI
• Challenges
• Managing shared restoration capacity
• Timely and reliable failure detection and
  notification
• Coordination between different layers in
  protection/restoration

28
Protection/Restoration in IP over Optical
Office B
RA
RC
RB1
RB2
Office B
RD
RE
RF

• Link failure Optical Layer protection/restoration
  
• Router failure IP rerouting gt need extra spare
  capacity
• gt planned w/ spare capacity for each single
  router failure
• gt The build-in spare capacity to cover any
  single router failure also covers single link
  failure gt no need to buy protected links with
  more than double the cost of unprotected links
• Long mean-time-to-repair for a failed link gt
  severe congestion if a router fails during an
  unprotected link failure

29
A New Joint IP/Optical Restoration Mechanism

• Objective more cost-effective and faster
  restoration in the IP over Optical architecture

30
A New Joint IP/Optical Restoration
MechanismMulti-Layer View

• Objectives
• Use Optical Layer Reconfiguration To Provide
  Restoration After A Backbone Router Failure
• Remove Need For IP Layer Restoration Capacity
  For Node Failures - Thereby Allowing Lower
  Optical Layer Unit Costs To Be Leveraged

RB1
RB2
RA
RC
OXCB
OXCA
OXCC
OXCB
OXCB
OXCB
RD
RE
RF
31
A New Joint IP/Optical Restoration MechanismIP
Layer Topology
RB1
RB2
RA
RC
Office B
RD
RE
RF

• Comparison with current IP rerouting
• For traffic to or from the office gt same of
  routing hops with no need for additional backbone
  capacity, need the same additional intra-office
  capacity as in IP rerouting
• For transit traffic gt shorter by 1 intra-office
  hop with no need for additional capacity
• Current IP rerouting needs additional backbone
  capacity for both cases

Reference A. Chiu, J. Strand, "Joint IP/Optical
Restoration After A Router Failure", OFC 2001