Label Switched Path (LSP) Dynamical Provisioning Performance

Download Report

Transcript Label Switched Path (LSP) Dynamical Provisioning Performance

Label Switched Path (LSP) Dynamical
Provisioning Performance Metrics in
Generalized MPLS Networks
draft-xie-ccamp-lsp-dppm-00.txt
Guowu Xie,SJTU
Guoying Zhang, CATR MII
Jianghui Han, IXIA
Xueqing Wei, Fiberhome
Jianhua Gao, Huawei
Motivations
• Why need for measuring and characterizing LSP
provisioning performance in GMPLS networks.
– Requirements of dynamical provisioning capability in GMPLS
Networks: interactive broadband multimedia services.
– Let the requirement of applications and the provisioning
capability of the network be mapped to each other
– Essential to the GMPLS network maintenance and
optimization.
• Standardized metrics and the way to measure them
precisely should be defined.
• The objective is to maximize common understanding of
LSP provisioning performance between the service
providers and network users.
Methodology
• GMPLS LSP performance measurement is similar to
IP network:
– GMPLS control plane is a packet switched network
– the LSP setup/release delay is similar to the round-trip Delay
in IP networks.
• We use the structures and notions introduced and
discussed in the IPPM framework document,
[RFC2330] [RFC2679] [RFC2681], and we define
the metrics according to GMPLS specific message
exchange scenarios.
Outlines
• Singleton definition of performance metrics
– Unidirectional LSP Setup Delay
– Bidirectional LSP Setup Delay
– LSP Graceful Release Delay
• Samples definition of performance metrics
– pseudo-random Poisson process is used
– Other models may be choosed
• Statistics Performance
– Min/Max/Median/Percentile
– Variance
Outlines
• Measuring these metrics is important for applications
scheduling.
– Longer LSP setup delay will incur higher overhead for the
requesting application.
• Also reflects the status of control plane.
– For LSPs traversing the same route, longer setup delays may
imply congestion in the control channel or high control element
load.
Real Network Demonstration
• China High Performance Broadband Information Network
(3TNET) ----863 project Sponsored by the Ministry of
Science and Technology.
– Circuit-switched and packet-switched hybrid network to support
large-scale interactive broadband streaming media.
– Network Infrastructure: IP +ASON/GMPLS+DWDM
– Services: VOD, DVB-IP, high speed Internet access
• ASON/GMPLS network has some enhanced features of
burst scheduling and multicasting, to provide dynamic and
automated provisioning of end-to-end Ethernet connections
to media server and IPTV head-ends.
– Dynamic point-to-multipoint connections for IP head-ends
– Periodic connection provisioning for media server, every 10 secs
Measured Performance in the Field Trial
500
1000
1500
2000
2500
500
1000
1500
2000
2500
99.999
1000
500ms/100ms
800
Setup Delay (ms)
Cumulative Counts
99.9
600
400
99
95
80
60
40
20
5
1
0.1
0.01
16000
200
0
0
100
200
14000
300
12000
Counts
Request Sequence
• Delay Variance observed
at any load. Higher load
will lead to higher
variance and longer
setup/release delay.
10000
8000
6000
4000
2000
0
Setup Delay (ms)
Average setup delay 668 ms
99.9% setup delay less than 1000ms
Open Issues
• time synchronization between the testing nodes
• other circumstances, such as point-tomultipoint multicast LSP, LSP rerouting
• Enhancement of MIBs: support monitoring of
LSP provisioning metrics on GMPLS nodes,
for maintenance/optimization.
• Expect to receive feedbacks and suggestions!