Transcript ppt

Cross-layer Visibility as a
Service
Ramana Rao Kompella
Albert Greenberg, Jennifer Rexford
Alex C. Snoeren, Jennifer Yates
1
Layering in the current Internet
OVERLAYS
MPLS
IP
Ethernet Optics
Fiber-spans
Fiber
2
Layering is a mixed blessing
 Layering allows us to contain complexity
 Each layer evolves independently without
affecting any other layer
 Allows us to focus on one layer at a time
 There are associated challenges too…
 Routine operational tasks need associations
across layers
 Example: mapping an IP link to optical circuit,
overlay link to an IP path
 Lack of accurate cross-layer associations can
affect the reliability of the network
3
Intended planned maintenance
Seattle
Planned
maintenance
on optics
Boston
Denver
San Francisco
Chicago
New York
St Louis
Los Angeles
Dallas
Orlando
4
Intended planned maintenance
 Optical component is
on circuit id A
LA to San
Traffic from LA to
 Lookup database to
Francisco link is
Dallas is rerouted
congested
map circuit id A to IP
via Denver
Planned maintenancelink
can
Denver
induce faults 
if Due to mis-association,
San Francisco
accurate associationsincorrectly
are
maps it to LA
not maintained to Dallas
Los Angeles
 Increase OSPF weight
Dallas
High
on LA to Dallas link
OSPF
 Disconnect component
weight
 Causes failure
X
5
Customer Fault Tolerance
New York
Internet
Shared optical
element
Customer diversity
Customer in NJ
information requires
INTRA-CARRIER
Philadelphia
accurate
cross-layer
DIVERSITY
Sprint
associations, sometimes
Level 3domains
across
New York
Internet
Going through
same conduit or
Holland tunnel ?
Customer in NJ
INTER-CARRIER
DIVERSITY
6
Fault diagnosis
Seattle
Because of a bug,
IP forwarding path
changed, but MPLS
did not !
Boston
Denver
San Francisco
Los Angeles
Chicago
Diagnosing faults
requires
accurate cross-layer
St Louis
associations
X
Dallas
New York
Orlando
MPLS circuit
between LA and
New York
What
happened
?!!
7
Why is it hard ?
 Can’t the operators maintain associations in a
centralized fashion ?
 Maintain database as links are provisioned
 Update as and when interfaces are re-homed
 Hard due to flux in topology
 Churn because of dynamic topology changes
 Human errors during re-homing interfaces
 Operational realities – separation of concerns
8
How it is done today ?
 A combination of non-standard databases
 Human-generated inventory data
 Measurement data obtained from probes
 Configuration state from network elements
 Policies implemented in network elements
 Higher complexity and overhead
 No compatibility across ASes
 Difficult to evolve a network
 Difficult to integrate two networks after acquisition
 Difficult to incorporate third-party tools
9
Why not concentrate on restoration?
 Advantages of lower-layer restoration
 Hides lower-layer failures from impacting upper
layers
 Obviates to some extent need for cross-layer
visibility
 Cross-layer visibility still important
 Lower-layer restoration more expensive than IP
restoration
 Subtle performance changes (e.g., RTT) need
diagnosis
10
Why not fatten the interfaces ?
 Fattening interfaces to make layers aware of
the entire topologies above and below
 Layers discover and propagate mappings
automatically
 Management system can query the network to
obtain mappings
 Fattening results in high complexity
 Interoperability is a big challenge – long
design and test cycles
 Wider interfaces impact security
11
Architecture for cross-layer visibility
BOW-TIE
OVERLAYS
Backbone
planning
MPLS
Cross-layer
Policy
Server
IP
Ethernet Optics
DB
Fiberspans
Ping
Trace-route
Customer
diversity
Backbone
maintenance
Fault
diagnosis
Fiber
IP
HOUR-GLASS
MANAGEMENT
APPLICATIONS
12
Standardize what goes in !
OVERLAYS
Standardize what goes in
(e.g. IP topologies)
MPLS
IP
Optics
FIBER, FIBERSPAN
Facilitates interaction
between ISP policy servers
AS1
OVERLAYS
MPLS
IP
Optics
FIBER, FIBERSPAN
AS2
13
Advantages of the bow-tie
 Topology, routing information and other
associations can be queried for maintenance,
diversity, and fault diagnosis
 Cooperation across ASes to present better
visibility across domains
 Policies easily enforced through the server
 Lower overhead on network elements
 Caching of common queries possible
 Historical questions can be answered
14
Evolution path to improve accuracy
 A lot of room for improvement
 Architecture accommodates evolution so that
accuracy can be improved over time
 Evolution path for individual layers
 Fiber & Fiber-spans
 Optical components
 IP links
 MPLS and overlay paths
15
Fiber & Fiberspans
 Automated mechanisms
[sebos02]
FIBER
GPS
OPTICAL
TAPS / RFID
DB
FIBER
 Inject labels through
fibers or use RFID
 GPS to determine the
location of fibers
 Transmit this information
to the DB
 More coverage results in
better accuracy but
expensive
16
Optical components
 Manual mechanisms
 Basic consistency checks
 Automatic correlation mechanisms such as
[kompella05nsdi] to output errors
 Automatic mechanisms
 Neighbor discovery for active optical devices
 Configuration state from “intelligent” optical
networks (that support dynamic restoration)
17
Optical components
Configuration state
during restoration
Neighbor discovery
through periodic
broadcasts at optical
layer
Intelligent
Optical
Network
ROUTER B
ROUTER A
DB
18
Other layers
 IP layer
 Periodically obtain configuration information to
construct topology
 Automatically collect up/down messages to
provide up-to-date view
 MPLS and overlay paths
 Static paths obtained from configuration
 Dynamic paths obtained by monitoring signaling
messages
19
Summary
 Accurate associations critical to many
operational tasks
 A bow-tie architecture for cross-layer visibility
 Provides the cross-layer associations as a service
to various applications
 Allows better cooperation among ASes through
standardizing what goes into the database
 Policy controlled export of these associations
 Lower overhead on network elements
 Allows for innovation while containing complexity
20
Future research directions
 Design automated mechanisms at each layer
to improve cross-layer visibility
 What frequency should information be obtained?
 How do we resolve conflicts (minimal edits) in the
database?
 Identify higher-level models that we need to
standardize
 Devise incentives for cooperation among
ASes
 Define a language to specify policies
21
Questions ?
22