Assessing the Performance of Multi- Layer Path Computation - - PowerPoint PPT Presentation

assessing the performance of multi layer path computation
SMART_READER_LITE
LIVE PREVIEW

Assessing the Performance of Multi- Layer Path Computation - - PowerPoint PPT Presentation

Assessing the Performance of Multi- Layer Path Computation Algorithms for different PCE Architectures, Anaheim, 2013, March 20 th , S. Martinez, V. Lpez, M. Chamania, . Gonzlez de Dios, A. Jukan and J. P. Fernndez-Palacios 0


slide-1
SLIDE 1

Telefónica I+D

Telefónica Servicios Audiovisuales S.A. / Telefónica España S.A. Título de la ponencia / Otros datos de interés / 26-01-2010 TELEFÓNICA I+D

Anaheim, 2013, March 20th,

  • S. Martinez, V. López, M. Chamania, Ó. González

de Dios, A. Jukan and J. P. Fernández-Palacios

Assessing the Performance of Multi- Layer Path Computation Algorithms for different PCE Architectures,

slide-2
SLIDE 2

Telefónica I+D

1

1 Telefónica Servicios Audiovisuales S.A. / Telefónica España S.A. Título de la ponencia / Otros datos de interés / 26-01-2010 TELEFÓNICA I+D

  • S. Martinez, V. López, M. Chamania, Ó. González de Dios,
  • A. Jukan and J. P. Fernández-Palacios

Anaheim, 2013, March 20th

Assessing the Performance of Multi- Layer Path Computation Algorithms for different PCE Architectures,

slide-3
SLIDE 3

Telefónica I+D

2

Index

Introduction Implemented ML Algorithms Results Conclusions and Next Steps

01 02 03 04

slide-4
SLIDE 4

Telefónica I+D

3

Introduction

01

Telefónica I+D

slide-5
SLIDE 5

Telefónica I+D

4

Introduction §

A typical core network is based on an IP/MPLS network designed over a transport network (e.g. Wavelength-Switched Optical Network).

§

Multi-layer interworking is considered to increase efficiency.

§

Joint computation (IP/MPLS+WSON) is key.

§

Path Computation Element: standard approach to solve the computation

§

Two PCE-based schemes proposed to compute multi-layer paths:

  • Coordinated (multiple) PCEs solution: each layer has its own PCE.
  • Integrated(single) multi-layer PCE where a unique PCE has the complete

multi-layer topology information.

§

We have implemented PCE-based schemes and several algorithms.

§

Performance is assessed in terms of blocking and computation delay

slide-6
SLIDE 6

Telefónica I+D

5

Cooperating PCEs scheme

slide-7
SLIDE 7

Telefónica I+D

Multilayer PCE scheme §

Single multilayer graph to compute all the requests.

6

slide-8
SLIDE 8

Telefónica I+D

7

Multi-layer Paths §

ERO: Standard representation of a path.

§

Multi-layer Path includes path in upper layer and paths in lower layer

§

Multi-layer ERO: standard RSVP ERO subobjects and use the SERVER_LAYER_INFO sub-object to indicate a layer change.

§

Virtual Network Topology Manager (VNTM) takes care of the establishment of lower layer paths

§

ERO inside a TE-Link Suggestion message is sent to VNTM.

slide-9
SLIDE 9

Telefónica I+D

Emulated Setup §

Multi-layer PCE emulates the 2 modes (integrated and cooperation).

§

Cooperation between PCEs demonstrated in Chamania et al. ICC 2012.

8

slide-10
SLIDE 10

Telefónica I+D

Message Exchange

9 The ¡NMS ¡starts ¡the ¡PCEP ¡ session ¡with ¡the ¡ML-­‑PCE

O P E N K e e p a l i v e O P E N K e e p a l i v e R e q u e s t ¡ ( B W )

The ¡NMS ¡Sends ¡a ¡ request ¡to ¡the ¡ML-­‑PCE

R e s p

  • n

s e ¡ ( M L

  • ­‑

E R O )

A

ERO ¡ computation ¡ time ¡(ms)

The ¡NMS ¡detects ¡that ¡it ¡is ¡a ¡ ML-­‑Response ¡and ¡suggest ¡the ¡ links ¡to ¡the ¡VNTM

T E _ L I N K _ S U G G E S T I O N ( M L

  • ­‑

E R O ¡ i n c l u d e d ) TE_LINK_CONFIRMATION (ML ¡ERO ¡included)

A

LSP ¡ establishment ¡ time ¡(ms)

OSPF ¡Link ¡Notification (Update ¡Optical ¡TEDB/graph) OSPF ¡Link ¡Notification (Update ¡Optical ¡TEDB/Graph)

VNTM NMS ¡(Client ¡PCC) ML-­‑PCE

slide-11
SLIDE 11

Telefónica I+D

10

Implemented ML Algorithms

02

Telefónica I+D

slide-12
SLIDE 12

Telefónica I+D

Implemented Algorithms: Two TEDBs Architecture §

Multi-layer path computation for a request in the client network can be reduced to a combination of one or more of four different operations:

  • Operation 1: Route the traffic onto an existing light-path directly connecting

the source and the destination .

  • Operation 2: Route the traffic through multiple existing light-paths.
  • Operation 3: Set up a new light-path directly between the source and the

destination and route the traffic onto this light-path.

  • Operation 4: Set up one or more light-paths that do not directly connect the

source and the destination , and route the traffic onto these light-paths and/

  • r some existing light-paths.

11

  • H. Zhu, et al, “A novel generic graph model for traffic grooming in heterogeneous WDM

mesh networks,” IEEE/ACM Trans. , 2003.

slide-13
SLIDE 13

Telefónica I+D

Implemented Algorithms : Two TEDBs Architecture §

The combination of the four operations leads us to the following algorithms for multilayer routing:

  • Minimizing the Number of Traffic Hops (MinTH).
  • Minimizing the Number of Light-paths (MinLP).
  • Minimizing the Number of Wavelength-Links (MinWL).

12

slide-14
SLIDE 14

Telefónica I+D

Algorithms Implemented: One TEDBs Architecture §

Multilayer auxiliary graph

§

Different weights to the edges in order to reach the different grooming policies by applying K-shortest-paths to this graph.

13

slide-15
SLIDE 15

Telefónica I+D

14

Results

03

Telefónica I+D

slide-16
SLIDE 16

Telefónica I+D

Validation scenario §

Telefonica core Network Scenario:

  • Client Layer: MPLS (PSC).
  • Transport Layer: Optical/Lambdas (LSC).

§ 80 lambdas/fiber.

§

Bandwidth requests:

§ 1-10 Gbps at MPLS layer. § Sequential requests. § Random order.

15

slide-17
SLIDE 17

Telefónica I+D

Evaluation of ML algorithms

16

1101 ¡con. 889 con. 1362 ¡con. 962 ¡con.

slide-18
SLIDE 18

Telefónica I+D

Evaluation of ML algorithms (II)

17

§

MinTH policy can improve total capacity allocation by 13.91% when using a single PCE as compared to multiple PCEs while the increase in the MinLP policy is 9,97%.

slide-19
SLIDE 19

Telefónica I+D

Computation time Evaluation §

Mean Computation time on Single ML PCE:

  • 54.32 ms MinLP, 53.57 ms MinTH

§

Multiple PCE solution:

  • 29.15 ms MinLP, 27.19 ms MinTH
  • Additionally, delay between PCEs per each query (0 in this setup)

§

Multi-layer PCE solution computation time bigger due to size of graph.

  • the PCE is running KSP in a large graph with N+1 layers (81 in this case).
  • KSP complexity increases linearly with number of arcs and nodes the

computation time for this algorithm is much higher than in the case of

§

Multiple PCE

  • Each PCEs runs their own KSP+First Fit.

18

slide-20
SLIDE 20

Telefónica I+D

Number of operations

19

MinTH MinLP MinWL 200 400 600 800 1000 1200 1400 Number Operations 1 TEDB # Operation 1 # Operation 2 # Operation 3 # Operation 4 MinTH MinLP MinWL 200 400 600 800 1000 1200 1400 Number Operations 2 TEDB # Operation 1 # Operation 2 # Operation 3 # Operation 4

§

In two TEDB case, interactions between PCEs is limited to one.

slide-21
SLIDE 21

Telefónica I+D

20

Conclusions and Next Steps

05

Telefónica I+D

slide-22
SLIDE 22

Telefónica I+D

Conclusions §

Multi-layer coordination is becoming an important requirement in network operators, due to its potential to reduce carrier’s investments, network interventions and thus cost of operations.

§

Multi-layer PCE is a key technology for multi-layer coordination and interworking.

§

This work shows the performance achieved with full topological information or with partial information.

§

Results show that integrated TED PCE can increase 13,91% bandwidth utilization in Spanish topology, but its computational time increases 49,24% in comparison with two coordinated TEDs per layer.

21

slide-23
SLIDE 23

Telefónica I+D

Algorithms Evaluation §

Use of resources in the network for the different algorithms in a non blocking static scenario (results for 400 requests):

22 MinLP MinTH MinWL 200 400 600 800 1000 1200 1400 Resource use 1 TEDB # LP # WL # Hops x 100 MinLP MinTH MinWL 200 400 600 800 1000 1200 1400 Resource use 2 TEDB # LP # WL # Hops x 100