network planning & design: an art or a science?
DESCRIPTION
With the rapid growth of IP networks in South-Asia in the past few years, and the advent of new services and applications -- be they wireless/wireline broadband Internet access, cable telephony, VoIP, remote teleconferencing, e-governance, or mobile entertainment -- a key issue before carriers is how to design and operate their networks as methodically and as efficiently as possible to maximize both customer retention and profits. While several best practices typically emerge from each provider\'s unique situation and cumulative experience (the "art" of network design), there are certain operational precepts that systematize and streamline the complex, multi-dimensional task of designing and managing modern, operational IP networks (the "science" of network design). In this talk, we first discuss the overall network design process and the manner in which control over the network must be exercised at varying timescales to achieve efficient operation. Next we discuss the functions that the operational, engineering, and planning teams at a carrier must typically execute, their inter-relationships, and the importance/rationale for performing them to optimize network performance. We then outline some network design best practices that have evolved over the past decade, drawing upon examples of carriers such as Sprint, Global Crossing, AT&T, NTT, and Reliance. We conclude with a look at some automated traffic engineering and planning tools, and how they enable carriers to rapidly identify potential performance problems, rigorously experiment with/evaluate design options, perform thorough scenario and network analysis, and develop robust designs.TRANSCRIPT
Network Planning and Design: Network Planning and Design: An Art or a Science?An Art or a Science?
Vishal Sharma, Ph.D.Metanoia, [email protected]://www.metanoia-inc.com
Metanoia, Inc.Critical Systems Thinking™
© Copyright 2007All Rights Reserved
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 2
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Agenda
Overall network planning process
Network design -- control at different timescales
“Science” of network planning/design Architecture and design philosophy
Network planning functions by organization – ops, eng., & planning
“Art” and “science” of network planning/design Best practice examples – GBLX, Sprint, Reliance
Role of automated TE and planning tools
Conclusions
Overall Network Design Overall Network Design ProcessProcess
Metanoia, Inc.Critical Systems Thinking™
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 4
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
A View of the Overall Network Planning/Layout Process
SS7 SignalingNetwork Design
Optical Link Configuration
Bandwidth Requirements
Link Requirements
Switch NetworkDesign
ATM NetworkDesign
IP/MPLS NetworkDesign
SONET/SDHNetwork Design
Optical RingNetwork Design
Optical MeshNetwork Design
Traffic DemandGeneration/Estimation
SONET/SDHSynchronization
We are here
Note: In a combined multi-layer network design, this strict division may not hold
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 5
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Network Design: Exercising Control over Different Timescales
Forecast traffic demandsTraffic requirements
Performance objectives
Traffic measurement dataObserved network statesForecast
Load
Load Uncertainties
Traffic Managementpkt. level processing,
routing control, congestionmanagement
Capacity Managementcapacity allocation, cap.planning, routing, design
management
Network Planning network dimensioning,modeling, perf. analysis,
what-if analysis
ms, sec, min
days, weeks
months, years
Controls(config/reconfig)
ActualLoad
Network
Datacollection
Network modeling, layout, &expansion process
The “Science” of Network The “Science” of Network PlanningPlanning
Metanoia, Inc.Critical Systems Thinking™
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 7
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Network Design System: Architecture
Centralized or distributed?
Obtaining network topology &
utilization info?
Inputs
Their formats?
Interface with n/w elements?
Outputs
Their formats?
Installing routes/LSPs in n/w?
Path route computation
On-line, dynamic?
Off-line, global?
Combination of above?
Route computation trigger
User? Administrator?
New request(s)?
Scalability
# of links & nodes handled?
# of flows, LSPs, circuits?
# of constraints handled?
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 8
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Network Design System: Philosophy
Control-centric
Long time scales
Large granularity flows
Verisimilitude-centric
Pkt. by pkt. sims. of elements & network
Traffic trace-based perf. simulations
Hybrids
Long time scale analysis for network perf.
Pkt. level sims. for element & flow performance
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 9
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Key Carrier Network Design Functions by Organization
Operations Engineering/Architecture Planning
Network design + opt.
Flow analysis
Simulation &experimentation
Expansion planning
Multi-protocolmodeling
Survivability analysis
Economic analysis
Monitoring
Diagnostics
Validation
1
2
3
4
5
6
78
9
10
The “Art” of Network Planning The “Art” of Network Planning (and some “science”)(and some “science”)
Metanoia, Inc.Critical Systems Thinking™
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 11
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Global Crossing IP Backbone Network
100,000 route miles 27 countries 250 major cities5 continents200+ POPs
Courtesy: Thomas Telkamp, GBLX
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 12
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Global Crossing: Network Design Philosophy
Ensure there are no bottlenecks in normal state
On handling congestion
Prevent via MPLS-TE
Manage via Diffserv
Over-provisioning
Well traffic engineered network can handle all traffic
Can withstand failure of even the most critical link(s)
Avoid excessive complexity & features
Makes the network unreliable/unstable
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 13
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Global Crossing’s Approach: Big Picture
WebServer
HR
DR BR
AR
CR
WR
DR
HR BR
AR
CR
WR
DR
HR BR
AR
CR
WR
EthernetSwitch
ModemBank
To other ISPs
To Customers
POP1
POP2
POP3
AR = Access Router
BR = Border Router
CR = Core Router
HR = Hosting Router
WR = WAN Router
DR = DSL Aggregation
OC-3/OC-12
OC-12/OC-48
OC-48/OC-192
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 14
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Global Crossing’s LSP Layout and Traffic Routing Methodology
Region 1 Region 2
Region 3
Region 4
POP1POP3
POP4
POP5POP2
Full LSP Meshin Core
Core LSP betweenWRs in POPs 1 & 5
Source
Destination
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 15
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
SprintLinkTM IP Backbone Network
19+ countries
30+ major intl. cities5 continents(reach S. America as well)
400+ POPs
Courtesy: Jeff Chaltas Sprint Public Relations
Represents connectivity only (not to scale)
110,000+ route miles (common with Sprint LD network)
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 16
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
SprintLinkTM IP Design Philosophy
Large networks exhibit arch., design & eng. (ADE) non-linearities not seen at smaller scales
Even small things can & do cause huge effects (amplification)
More simultaneous events mean greater likelihood of interaction (coupling)
Simplicity Principle: simple n/wks are easier to operate & scale
Complexity prohibits efficient scaling, driving up CAPEX and OPEX!
Confine intelligence at edges
No state in the network core/backbone
Fastest forwarding of packets in core
Ensure packets encounter minimal queueing
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 17
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
SprintLinkTM Deployment StrategyL2 failure detection triggersswitchover before L3 converges
ZA
Parallel links 50% utilizationunder normal state
1
2
3
4
SONET framing forerror detection
LineCard
LineCard
SONETOverheadIP Data
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 18
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Reliance Communications: India Network
80,000-100,000 route kms
4000 towns, 350, 000 villages
3+ p2p paths on metro rings
Ring + mesh design, 130+ rings
Source: FLAG Telecom Extranet
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 19
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Reliance Network Design Philosophy Be integrated telecom provider: e2e ownership of entire architecture
Access systems, national fiber-optic backbone, international cable systems, Internet gateways, wireless base stations, and aggregation devices
80,000 km of high-capacity, pan-India fiber-optic digital network
Common infrastructure for wireline & wireless traffic Integrate packetized voice, video, data and circuit-switched voice on same
network
Ring architecture for physical redundancy (2+ disjoint paths)
SONET/SDH for framing and OAM capability (easier management)
MPLS LSPs for resilience (avoid SONET/SDH protection at PHY layer)
A Look at Planning ToolsA Look at Planning Tools(advancing the “science”)(advancing the “science”)
Metanoia, Inc.Critical Systems Thinking™
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 21
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Canonical Structure of Automated Planning Tools
H/w DeviceModels
User Interface/GUI/Command Line
Simulate
Analyze
Design
InternalEngine
ModulesIGP
MPLS-TE
Protection
RerouteOutputs
Config.Files
Inputs
Inputs
(Topo, Cap., Failure)Analysis
Reports
Eco. AnalysisForecasting
NMS
EMS EMS
Vendor Data Extraction
EquipmentInventory
Network
Data ExtractionData Conversion
Element DataElement State/Stats
Measurements/Probes
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 23
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
OPNET SP Guru: Components & Features
ProcessInputs
Discovery via NMplatforms for topology
Simulate failure &overload to get:- Worst-case loading- % disrupted traffic- SLA violation
Config.Verification
FlowAnalysis
FailureAnalysis
NetworkDesign
NetworkSimulation
MPLSModule
Integrated topo.,config., util. datafrom VNE Server
Switch/routerconfig. filesTraffic & link util.
data from systems
Traffic matrix info. fromdata collection tools
- Family of design algos.- TE capability- Impact of new algos. ortechnologies, e.g. VPNs
- “What-if” scenarios- Reachability analysis- Arch. & cost for givenQos, delay, jitter, loss- Current n/w loading
- Distribution of trafficon network links
Network model: nodes, links,speeds, topology, metrics
Validate configs.Deploy validated config.
Export n/wconfig.
- Primary/sec. path layout- FRR backup tunnels
Network simulation and analysis
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 25
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
OPNET Solutions: Key Characteristics
NetDoctor for configuration & network operations analysis
VNE Server
Automated I/F to various network data components
Ability to build a complete network view
Hybrid simulation techniques
Provide balance between speed and resolution
Ability to add SP’s own rules, algorithms, modules
Facility to map actual IP addressing to internal network model
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 26
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Cariden MATE: TE Process
DataCollection
DemandEstimation
RoutingOptimization
ChangeoverPlanning
User-definedConstraints
ChangeoverExecution
UserMonitoring
MATE’s TE System Components
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 27
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Cariden MATE: IGP Traffic Engineering
Problems
Uneven link utilization
Heuristic/ad-hoc planning
Coarse capacity upgrade rules (e.g. at 50 or 75%)
Above 60% utilization expected in 6 mo!
Sample network with projected traffic growth in 6 months
Original network state
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 28
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Cariden MATE: Routing Optimization
[Reproduced with permission: Cariden Technologies]
Objectives
Max. headroom on failure
Max. normal headroom
Minimize latency
Constraints
Fixed intra-site metrics
Symmetric weights
Latency bounds
Results
Max. link util. 89% 59%
Max. link util. on failure 110% 92%
All links brought to below 60% utilization
under normal conditions
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 29
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Cariden MATE: Resilience Capabilities
Before (worst case) After (worst case)
>> 95% utilization
Max. utilization under 92%
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 30
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Cariden MATE: Key Characteristics
Demand estimation & characterization Estimate p2p demands from aggregate node/intf. demands & routing
Estimate effective b/w per queueing class to meet QoS for demands
Robust routing changeovers Sequence of moves to transition network from one routing/LSP pattern to
another via a series of “make-before-break” operations
IGP metric-tuning based optimization in IP networks
Practical BGP simulations – peering, load balancing
Fully cross-platform – supports client/server or client-only model
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 31
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
A Perspective on Some Current Carrier Requirements
Accurate planning models
IP n/w planning with peering
Usability and consistency
Obtain precise traffic matrices
Good interface with monitoring tools
Intelligent heuristics
Extensible architecture
Application-level performance monitoring
Multicasting for multi-media services support
Keynote Talk, SANOG 9, Colombo, Sri Lanka, 23 January 2007 32
Metanoia, Inc.Critical Systems Thinking™
Copyright 2007All Rights Reserved
Concluding Remarks Designing and operating networks methodically will be crucial for
South-Asian carriers with growing networks
Services, not size, will dictate the sophistication needed
“Small” carriers offering same advanced services as “big” ones So, a systematic approach is must for sustained competitiveness
A “science” of network design can be used to … Identify potential problems, experiment with solns, verify designs
An “art” will always be there … Dictated by unique local or evolutionary needs
Tools and expertise exist to help operate world-class networks!