[PPT] - ViAggre: Making Routers Last Longer! Hitesh Ballani Paul Francis, PowerPoint Presentation

SLIDE 1

ViAggre: Making Routers Last Longer!

Hitesh Ballani Paul Francis, Tuan Cao and Jia Wang Cornell University and AT&T Labs – Research HotNets 2008

SLIDE 2

Motivation: Rapid Routing Table Growth

50000 100000 150000 200000 250000 300000 08 06 04 02 00 98 96 94 92 90 88 Active BGP entries (FIB Size) Year 282,000 prefixes (Sep’08) 50000 100000 150000 200000 250000 300000 08 06 04 02 00 98 96 94 92 90 88 Active BGP entries (FIB Size) Year 282,000 prefixes (Sep’08)

SLIDE 3

Motivation: Rapid Routing Table Growth

100000 200000 300000 400000 500000 14 08 06 04 02 00 98 96 94 92 90 88 Active BGP entries (FIB Size) Year 100000 200000 300000 400000 500000 14 08 06 04 02 00 98 96 94 92 90 88 Active BGP entries (FIB Size) Year

??

Rapid future growth

◮ IPv4 exhaustion ◮ IPv6 deployment

SLIDE 4

Typical Router Innards

Route Processor CPU RIB Line Card ASIC FIB Line Card Line Card Line Card

SLIDE 5

Typical Router Innards

Route Processor CPU RIB Line Card ASIC FIB Line Card Line Card Line Card

Routing Information Base (DRAM $)

SLIDE 6

Typical Router Innards

Route Processor CPU RIB Line Card ASIC FIB Line Card Line Card Line Card

Forwarding Information Base (SRAM $$$)

SLIDE 7

Does (FIB) Size Matter?

Technical concerns

◮ More Memory ◮ More Processing ◮ Power and Heat dissipation problems

SLIDE 8

Does (FIB) Size Matter?

Technical concerns

◮ More Memory ◮ More Processing ◮ Power and Heat dissipation problems

Business concerns

◮ Less cost-effective networks

◮ Price per byte forwarded increases

◮ Router memory upgrades

SLIDE 9

Does (FIB) Size Matter?

Technical concerns

◮ More Memory ◮ More Processing ◮ Power and Heat dissipation problems

Business concerns

◮ Less cost-effective networks

◮ Price per byte forwarded increases

◮ Router memory upgrades

ISPs are willing to undergo some pain to reduce FIB size

SLIDE 10

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 11

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

Separate edge from the core

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 12

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

Geographical routing

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 13

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

Compact routing

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 14

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

All require architectural change So many ideas, so little impact!

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 15

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

Tackle routing scalability through a series of incremental, individually cost-effective upgrades

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 16

Routing Scalability Problem Space

[Deering, ’96] [O’Dell, ID’97] [Zhang et. al., ICNP’06] [Farinacci, ID’07] [Massey et. al., ID’07] [Jen et. al., HotNets’08] [Francis, CNIS’94] [Deering et. al., ID’00] [Hain, ID’02] [Krioukov et. al., Arxiv’05]

This Paper: Focuses on reducing FIB size

FIB growth RIB growth Routing Convergence, Update Churn, ....

SLIDE 17

Virtual Aggregation, aka ViAggre

A “configuration-only” approach to shrinking router FIBs

◮ Applies to legacy routers ◮ Can be adopted independently by any ISP

Key Insight: Divide the routing burden A router only needs to keep routes for a fraction of the address space

SLIDE 18

Talk Outline

◮ Introduction[]y ◮ ViAggre: Basic Idea[]y ◮ ViAggre Design[]y ◮ Evaluation[]y ◮ Deployment[]y ◮ Conclusions

SLIDE 19

ViAggre: Basic Idea

PoP A PoP C PoP B

ISP

IPv4 Address Space 0.0.0.0 255.255.255.255 External Router External Router

Today: All routers have routes to all destinations

SLIDE 20

ViAggre: Basic Idea

PoP A PoP C PoP B

ISP

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Virtual Prefixes

Divide address space into Virtual Prefixes (VPs)

SLIDE 21

ViAggre: Basic Idea

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Virtual Prefixes

Aggregation Points for Green VP

Assign Virtual Prefixes to the routers Routers only have routes to a fraction of the address space

SLIDE 22

ViAggre: Basic Idea

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Virtual Prefixes

Aggregation Points for Green VP

How to achieve such division of the routing table? Without changes to routers and routing protocols Without cooperation from external networks

SLIDE 23

Talk Outline

◮ Introduction[]y ◮ ViAggre: Basic Idea[]y ◮ ViAggre Design[]y ◮ Evaluation[]y ◮ Deployment[]y ◮ Conclusions

SLIDE 24

ViAggre Control-Plane

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Full Routing Table

eBGP Peers may advertise full routing table

SLIDE 25

ViAggre Control-Plane

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Full Routing Table Load full routing table into RIB Supress all but blue routes from FIB

FIB Suppression Blue routers only load blue routes into their FIB

SLIDE 26

Data-Plane paths

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

Virtual Prefix Packets destined to a prefix in Red

Consider packets destined to a prefix in the red VP

SLIDE 27

Data-Plane paths

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1 2

ViAggre path Ingress (I) → Aggregation Pt (A) → Egress (E)

SLIDE 28

Ingress → Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1

Router I doesn’t have a route for destination prefix

SLIDE 29

Ingress → Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1 Advertise Red VP

Aggregation Points advertise corresponding Virtual Prefixes

SLIDE 30

Ingress → Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1 Advertise Red VP

Prefix Next-Hop P1 P2 .... .... 0/2 A 128/2 192/2 .... ....

Blue router has a route for the red Virtual Prefix

SLIDE 31

Aggregation Point → Egress

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

2

Prefix Next-Hop P3 P4 X .... 64/2 .... 128/2 192/2 .... ....

Aggregation Pt. A tunnels packet to external router

SLIDE 32

Aggregation Point → Egress

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1 Strip tunnel header from outgoing pkts

Egress Router strips the tunnel header off outgoing packets

SLIDE 33

Failure of Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1

What if Aggregation Pt. A fails?

SLIDE 34

Failure of Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

Prefix Next-Hop P1 P2 .... .... 0/2 A2 128/2 192/2 .... ....

Router I installs the route advertised by A2

SLIDE 35

Failure of Aggregation Point

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

Prefix Next-Hop P1 P2 .... .... 0/2 A2 128/2 192/2 .... ....

Packets are re-routed appropriately

SLIDE 36

ViAggre’s impact on ISP’s traffic

0.0.0.0 255.255.255.255 External Router External Router 0/2 64/2 128/2 192/2

I A E X A2

1

ViAggre paths can be longer than native paths Traffic stretch, increased router and link load, etc.

SLIDE 37

Popular Prefixes

Traffic volume follows power-law distribution

◮ 95% of the traffic goes to 5% of prefixes ◮ Has held up for years

Install “Popular Prefixes” in routers

◮ Stable over weeks ◮ Mitigates ViAggre’s impact on the ISP’s traffic

SLIDE 38

Talk Outline

◮ Introduction[]y ◮ ViAggre: Basic Idea[]y ◮ ViAggre Design[]y ◮ Evaluation[]y ◮ Deployment[]y ◮ Conclusions

SLIDE 39

Stretch Vs FIB Size

Assigning more routers to aggregate a virtual prefix

◮ Reduces Stretch imposed on Traffic ◮ Increases FIB size

Aggregation Point Assignment Problem

◮ Minimize Worst FIB size, subject to constraint

n Worst stretch

◮ NP-complete problem ◮ Implemented a greedy approximation

SLIDE 40

Performance Study

Data from tier-1 ISP

◮ Topology, Routing tables, Traffic matrix

Used out algorithm with varying stretch constraints

SLIDE 41

Constraining Worst Stretch

5 10 15 20 25 30 2 4 6 8 10 1 2 3 4 5 6 7 FIB Size (% of DFZ routing table) Stretch (msec) Worst-case Stretch Constraint (msec) Worst-case FIB Size 5 10 15 20 25 30 2 4 6 8 10 1 2 3 4 5 6 7 FIB Size (% of DFZ routing table) Stretch (msec) Worst-case Stretch Constraint (msec) Worst-case FIB Size

FIB Size reduces as Stretch constraint is relaxed

Worst-case Stretch ≤ 4ms ⇒ Worst FIB = 10,226 prefixes (4% of global routing table)

SLIDE 42

Constraining Worst Stretch

5 10 15 20 25 30 2 4 6 8 10 1 2 3 4 5 6 7 FIB Size (% of DFZ routing table) Stretch (msec) Worst-case Stretch Constraint (msec) Worst-case FIB Size Worst Stretch Average Stretch 5 10 15 20 25 30 2 4 6 8 10 1 2 3 4 5 6 7 FIB Size (% of DFZ routing table) Stretch (msec) Worst-case Stretch Constraint (msec) Worst-case FIB Size Worst Stretch Average Stretch

Average Stretch is negligible

Worst-case Stretch ≤ 4ms ⇒ Average Stretch = 0.2msec

SLIDE 43

Router Load

Deployment with Worst-case Stretch ≤ 4msec

◮ Shrinks FIB by more than 20x ◮ Median router load increases by 31.3%

Using popular prefixes

◮ 5% popular prefixes carry 96.7% of traffic ◮ Median and Worst-case router load increase ≈

1%

SLIDE 44

Talk Outline

◮ Introduction[]y ◮ ViAggre: Basic Idea[]y ◮ ViAggre Design[]y ◮ Evaluation[]y ◮ Deployment[]y ◮ Conclusions[]y

SLIDE 45

ViAggre Pros

◮ Shrinks router FIB substantially ◮ Can be incrementally deployed ◮ Can be deployed on a limited-scale ◮ Incentive for deployment ◮ No change to ISP’s routing setup

◮ Does not affect convergence times ◮ Does not affect routes advertised to

neighbors

◮ Does not restrict routing policies ◮ . . .

SLIDE 46

Can it be deployed?

Configuration overhead of a configuration-only solution

◮ Configuring FIB suppression on routers ◮ Configuring LSP advertisements on edge routers

Planning Overhead

◮ Choosing virtual prefixes ◮ Assigning aggregation points ◮ Assuring network robustness ◮ . . .

SLIDE 47

ViAggre management overhead

Deployed ViAggre on WAIL

◮ Cisco 7300 routers ◮ Developed Configuration Tool

◮ ∼330 line python script ◮ Extracts information from existing configuration files ◮ Generates ViAggre configuration files

◮ Planning tool in the works

Working with a router vendor (Huawei)

◮ Implement ViAggre natively ◮ IETF Draft

SLIDE 48

Conclusion

ViAggre shrinks the FIB on routers

◮ Can extend the lifetime of installed routers

Is this a “complete” solution? No

◮ A simple and effective first step ◮ Next Step: Inter-domain ViAggre

SLIDE 49