Table of Contents
Deploying Tight-SLA services on an IP Backbone
Objective
An overview of the Analysis
Further information
Agenda
Typical Core Per Class SLA Characteristics
Characterizing an IP Service
One-Way Latency
One-Way Jitter
Backbone VoIP Jitter Budget
One-Way Loss Rate
SLA Metrics: Throughput
TCP Throughput
SLA Metrics: Availability
Per flow sequence preservation
Re-ordering Impact on Service
Loss of Connectivity / Convergence
How to specify the target for the metric
Optimizing the IP Infrastructure
Agenda
Loss of Connectivity
Historical ISIS Convergence
What this presentation will explain
Link-State protocol overview
An example network
The Final SPT rooted at A
PPT Slide
The RIB construction
LSDB, RIB and FIB
SPF optimisations
SPF Optimizations
PRC
Incremental-SPF
Incremental-SPF
Incremental-SPF
Incremental-SPF
Incremental-SPF
SPF, PRC, I-SPF: summary
Topology and Leaf Optimizations
Parallel point-to-point adjacencies
P2P mode for back-to-back GE
Speeding up route installation
SPF, PRC and LSP-gen Exponential BackOff Timers
Backoff timer algorithm
Backoff timer algorithm
spf-interval 10 100 1000
Default Values
Two-Way Connectivity Check
Timers for Fast Convergence
Timer for Fast Convergence
LSP-Gen and Link Flap
LSP-Gen needs Intf Dampening
LSP Pacing and Flooding
LSP Pacing and Flooding
Link Protocol Properties
Link Protocol Properties
POS – Detection of a link failure
POS – Detection of a link failure
POS – Detection of a link failure
POS – Detection of a link failure
POS – Detection of a link failure
POS – Detection of a link failure
POS – Bringing a down link back up
POS – Best for Convergence
Other types of Links
Fast Hello’s
Operating this Design
Monitoring link/adjacency status
Before reloading a router
ISIS Monitor
ISIS Fast Convergence Design
Design Tips
Test Results
Test Scenari
PPT Slide
PPT Slide
PPT Slide
PPT Slide
SPF Duration
Down
Carrier-Delay
PPT Slide
Convergence
Conclusion
Conclusion
Agenda
OverProvisioned Backbone
The Key is OverProvisioning Offer must be higher than Demand
Over-Provisioned Backbone
Over-provisioning (Source: Stephen Casner, Packet Design, NANOG 22)
Drawback
Provisioning failure
“Not every week is like this” (Source: Stephen Casner, Packet Design, NANOG 22)
Recommendation: use DiffServ!
Service Isolation
Backbone Diffserv Design
Backbone Diffserv Design
Aggr Over-provisioning is Expensive
WRED Tuning
WRED Tuning
WRED Tuning
WRED Tuning
WRED Tuning
WRED Tuning
Typical Backbone Diffserv Design
Provisioning is simple
Capacity Planning
Mature Technology
EF: jitter charecteristic
EF: jitter characteristic
AF: Bandwidth Allocation Accuracy
AF: Latency = f(load)
Optimised for even rare/corner-cases
Diffserv Support over MPLS
MPLS and DiffServ Using the EXP bits
MPLS and Diffserv
Capacity Planning and Monitoring
Capacity Planning and Monitoring
Capacity Planning and Monitoring
Link statistics
Core Traffic Matrix
Active SLA Monitoring
IPPM Infrastructure
MPLS-based Technologies TE, DS-TE, FRR
TE and SLA’s
When TE is justified
MPLS FRR
Tight-SLA IP Backbone Conclusion
An overview of the Analysis
References
References
References
|