Transparent SDN Update Scheduling Peter Pereni @ EPFL Maciej - - PowerPoint PPT Presentation

transparent sdn update scheduling
SMART_READER_LITE
LIVE PREVIEW

Transparent SDN Update Scheduling Peter Pereni @ EPFL Maciej - - PowerPoint PPT Presentation

ESPRES: Transparent SDN Update Scheduling Peter Pereni @ EPFL Maciej Kuzniar @ EPFL Marco Canini @ UCLouvain Dejan Kosti @ KTH Network events trigger big updates Topology changes Network events trigger big updates Topology changes


slide-1
SLIDE 1

ESPRES: Transparent SDN Update Scheduling

Peter Perešíni @ EPFL

Maciej Kuzniar @ EPFL Marco Canini @ UCLouvain Dejan Kostić @ KTH

slide-2
SLIDE 2

Network events trigger big updates

Topology changes

slide-3
SLIDE 3

Network events trigger big updates

Topology changes

slide-4
SLIDE 4

Network events trigger big updates

Topology changes

slide-5
SLIDE 5

Network events trigger big updates

Topology changes

slide-6
SLIDE 6

Network events trigger big updates

Topology changes Traffic Engineering Policy changes

slide-7
SLIDE 7

Network events trigger big updates

Topology changes Traffic Engineering Policy changes

Many rule modifications ⇨ updates take time

slide-8
SLIDE 8

What if we reorder installations...

Update touching two (independent) flows

  • blue - 3 changes on a switch
  • red - 2 changes on a switch
slide-9
SLIDE 9

What if we reorder installations...

Update touching two (independent) flows

  • blue - 3 changes on a switch
  • red - 2 changes on a switch
slide-10
SLIDE 10

What if we reorder installations...

Update touching two (independent) flows

  • blue - 3 changes on a switch
  • red - 2 changes on a switch
slide-11
SLIDE 11

What if we reorder installations...

Update touching two (independent) flows

  • blue - 3 changes on a switch
  • red - 2 changes on a switch

vs

slide-12
SLIDE 12

What if we reorder installations...

Update touching two (independent) flows

  • blue - 3 changes on a switch
  • red - 2 changes on a switch

vs

total time same but different ordering of rule installation matters

slide-13
SLIDE 13

Some challenges

Scheduling sounds easy but

  • multiple switches affected
  • and switch-speeds are different over time
  • rule dependencies
  • e.g. update ingress only after core installed
  • control channel is FIFO
  • no reordering
slide-14
SLIDE 14

ESPRES overview

(See paper for more details)

  • Keep backlog of rule installations in

ESPRES

  • enables re-ordering on control channel
  • Schedule rules to be installed next
  • react on-the-fly to current switch conditions
  • needs to be fast
  • support flexible goals
slide-15
SLIDE 15

Possible scheduling goals

Already illustrated:

  • install majority of flows sooner

See paper:

  • decrease mid-update rule overhead
  • decrease transient instabilities
slide-16
SLIDE 16

1000 new flows; 18 switches

A taste of results

slide-17
SLIDE 17

1000 new flows; 18 switches

A taste of results

➢ No scheduling (send rule when dependencies are met)

slide-18
SLIDE 18

1000 new flows; 18 switches

A taste of results

➢ Espres A ➢ Espres B ➢ Optimal (Offline) ➢ No scheduling (send rule when dependencies are met)

slide-19
SLIDE 19

1000 new flows; 18 switches

A taste of results

➢ Espres A ➢ Espres B ➢ Optimal (Offline)

We reduce completion time by ≥40% for half of flows

➢ No scheduling (send rule when dependencies are met)

slide-20
SLIDE 20

Summary

  • Updates touch many flows
  • Changing rule installation order helps

achieve different goals ESPRES ➢Maintain backlog of rule installations at the controller ➢Schedule their order on-the-fly