Quantifying the Effectiveness of First-Hop Redundancy Protocols in - - PowerPoint PPT Presentation

Click to edit Master title style

Quantifying the Effectiveness of First-Hop Redundancy Protocols in IP Networks

Paul Bourne Neville Palmer & Jan S krabala

20 August 2018

Click to edit Master title style Presentation Outline

• S

takeholders

• Background
• Investigation
• method
• results and discussion
• recommendations and conclusions
• Questions and Answers

Click to edit Master title style

Hierarchical Enterprise Campus Design Architecture

(Cisco, 2008)

High Availability Networks

First Hop Redundancy Protocols

(Cisco, 2018)

Click to edit Master title style

• Metered secondary link
• without load balancing
• failover based on tracking obj ect
• HS

RP: Hot S tandby Routing Protocol

• GLBP: Gateway Load Balancing Protocol
• VRRP: Virtual Router Redundancy Protocol
• enable NTP and debug logging
• shut primary interface G0/ 0
• monitor line protocol and FHRP
• also 100 ms continuous PING

Method

Click to edit Master title style

• Metered secondary link
• without load balancing
• failover based on tracking obj ect
• HS

RP: Hot S tandby Routing Protocol

• GLBP: Gateway Load Balancing Protocol
• VRRP: Virtual Router Redundancy Protocol
• enable NTP and debug logging
• shut primary interface G0/ 0
• monitor line protocol and FHRP
• also 100 ms continuous PING

Method

Hello Interval Hold Timer Preemption Delay Priority

HSRP D

3 10

• HSRP R

1 3

• GLBP D

3 10 30

• GLBP R

1 3 3

• VRRP D

1

• 100/110

VRRP R

1

• 253/254

Click to edit Master title style

• VRRP fastest in failover and restoration with both configurations
• followed by HS

RP and GLBP wit h default configuration

• followed by GLBP and HS

RP wit h rapid configurat ion

• GLBP largely affect ed by 30 s default pre-empt ion delay
• HS

RP took longer to failover with the rapid configuration

• also slow t o rest ore links
• t he curious case of t he PING

Results and Discussion

Min Max Mean Dropped HSRP D

03.390 05.643 04.780 55

HSRP R

03.818 05.561 05.067 55

GLBP D

32.456 34.392 33.339 55

GLBP R

03.604 04.020 03.858 55

VRRP D

02.757 03.389 03.112 55

VRRP R

02.316 02.969 02.665 55

Min Max Mean Dropped HSRP D

25.133 34.931 29.635

HSRP R

25.312 29.521 27.876

GLBP D

32.507 36.524 34.253

GLBP R

03.284 04.295 04.040

VRRP D

02.372 03.363 02.989

VRRP R

02.236 02.970 02.583

Mean Failover Times Mean Restoration Times

Click to edit Master title style

• VRRP fastest in failover and restoration with both configurations
• followed by HS

RP and GLBP wit h default configuration

• followed by GLBP and HS

RP wit h rapid configurat ion

• GLBP largely affect ed by 30 s default pre-empt ion delay
• HS

RP took longer to failover with the rapid configuration

• also slow t o rest ore links
• t he curious case of t he PING

Results and Discussion

Min Max Mean Dropped HSRP D

03.390 05.643 04.780 55

HSRP R

03.818 05.561 05.067 55

GLBP D

32.456 34.392 33.339 55

GLBP R

03.604 04.020 03.858 55

VRRP D

02.757 03.389 03.112 55

VRRP R

02.316 02.969 02.665 55

Min Max Mean Dropped HSRP D

25.133 34.931 29.635

HSRP R

25.312 29.521 27.876

GLBP D

32.507 36.524 34.253

GLBP R

03.284 04.295 04.040

VRRP D

02.372 03.363 02.989

VRRP R

02.236 02.970 02.583

Mean Failover Times Mean Restoration Times

Click to edit Master title style

• VRRP appears to be the most responsive protocol
• also has more flexibilit y
• and is an open st andard
• VRRP removed authentication, which may cause concern
• should be implement ed wit hin a closed part of t he net work behind firewall and ACL
• alt ernat ively use FHRP where load balancing is not required
• or GLBP where load balancing is needed but adj ust t he t imers t o opt imise
• Low hold and pre-emption delays should be set with consideration for

flapping connections and boot time of devices – measure both to inform design

Recommendations and Conclusions

Click to edit Master title style

• Investigation raised several queries:
• consider rout ing prot ocol configurat ion and it s int eract ion wit h FHRP
• reduce t imers furt her t o find t he minimum failover t ime using msec
• invest igat e t he logging process t o det ermine t he accuracy of t imest amps
• monit or packet flows t o verify validit y of t imest amps
• int roduce load balancing and mult iple st andby devices
• Confident that the method produced valid relative results
• not convinced by t he accuracy of t he absolut e values recorded
• Compare observed behaviour with packet simulators: Packet Tracer, GNS

3 or Riverbed Modeller

• simulat ors rely on an accurat e implement at ion of S

W and modelling of HW

• use t o predict result s and discount null hypot hesis
• find opt imal configurat ions: prot ocol overhead, st abilit y and performance

Recommendations and Conclusions

Click to edit Master title style

Paul Bourne paul.bourne@ solent .ac.uk Neville Palmer neville.palmer@ solent .ac.uk Jan S krabala j an@ vost ron.com

Questions and Discussion

References: Cisco 2008. Campus Network for High Availability Design Guide [accessed 15/05/18]. Available: https://www.cisco.com/c/en/us/td/docs/solutions/Enterprise/Campus/HA_campus_DG/hacampusdg.html Cisco, 2018. First Hop Redundancy Protocols Configuration Guide [accessed 15/05/18]. Available: https://www.cisco.com/c/en/us/td/docs/ios-xml/ios/ipapp_fhrp/configuration/xe-16/fhp-xe-16-book/fhp- hsrp-mgo.html