FIB Aggregation

19
FIB Aggregation Zartash Uzmi draft-uzmi-smalta-01 (with Ahsan Tariq and Paul Francis)

description

FIB Aggregation. Zartash Uzmi draft-uzmi-smalta-01 (with Ahsan Tariq and Paul Francis). FIB Aggregation Work. Normal Router Operation. First introduced: IETF 76 draft-zhang-fibaggregation Level 1-4 SMALTA (at IETF78) Better (near-optimal). RIB Primary Routes. FIB. With Aggregation. - PowerPoint PPT Presentation

Transcript of FIB Aggregation

Page 1: FIB Aggregation

FIB Aggregation

Zartash Uzmi

draft-uzmi-smalta-01(with Ahsan Tariq and Paul Francis)

Page 2: FIB Aggregation

First introduced: IETF 76 draft-zhang-fibaggregation Level 1-4

SMALTA (at IETF78) Better (near-optimal)

FIB Aggregation Work

04/20/23 2FIB Aggregation

RIB

Primary Routes

FIB

Aggregate

FIB

Normal Router Operation

With Aggregation

Primary Routesfrom RIB

Page 3: FIB Aggregation

Completed, but not reflected in current draft Refinement of SMALTA Thorough Evaluation (with data from a real ISP)

High confidence level in results

In progress (Consolidation of the two drafts) Original (Level 1-4) draft (IETF 76) SMALTA draft (IETF 78)

Changes since IETF 78

04/20/23 3FIB Aggregation

Page 4: FIB Aggregation

Data Sets Routeviews (yearly: 12/2001 to 12/2010) Various routers from a Tier-1 service provider

Based on router type, location, #interfaces

Main findings: Savings In FIB memory (line card): 35% and upwards (as large as

75%) In #prefixes: ~12% better (than savings in memory) In lookup time (#memory accesses): ~25% faster Update processing: <1 FIB update per RIB update (on

average)

Evaluation of SMALTA

04/20/23 FIB Aggregation 4

Page 5: FIB Aggregation

L1/L2/SMALTA: Expectations?

04/20/23 FIB Aggregation 5

Router

SMALTA

Level 1 Level 2

R1 37% 68% 53%

R2 36% 66% 51%

R3 40% 68% 58%

R4 21% 55% 37%

R5 13% 49% 28%

R6 19% 54% 35%

R7 55% 79% 72%

For 2 Internet Gateway Routers (R1,R2) and 5 Access Routers in

Provider Network

Aggregated prefixes (as % of original)

Page 6: FIB Aggregation

Aggregation and #next hops

04/20/23 FIB Aggregation 6

Fewer aggregation opportunities with more nexthops

# Unique nexthops

% o

f Ori

gin

al

Routeviews

12/2010

Memory savings (for Tree Bitmap) are somewhat (~12%) lower

Page 7: FIB Aggregation

#Memory Accesses/Lookup time

04/20/23 FIB Aggregation 7

Lookup time (Tree Bitmap) varies in accordance with …the #prefixes after aggregation

Effective # Unique nexthops

5 Access Routers

(Provider Network)

% o

f Ori

gin

al

For Internet Gateway Routers, about 25% fewer memory accesses when using Tree

Bitmap

Page 8: FIB Aggregation

Incorporating Updates

04/20/23 FIB Aggregation 8

An IGR (Internet Gateway Router)

inProvider Network

12-hour Update Trace

% c

han

ge in

unaggre

gate

d

% o

f Ori

gin

al

#updates (in 10K)#aggregated prefixes is near-optimal after

a large number of updates are incorporated

Call toSnapsh

otfunction

Page 9: FIB Aggregation

Updates FIB downloads

04/20/23 FIB Aggregation 9

#Updates b/w consecutivecalls to “snapshot”

function

#Updates b/w consecutivecalls to “snapshot”

function

FIB

dow

nlo

ads

per

call

to

snap

shot

Cum

ula

tive F

IB

dow

nlo

ads

An IGR in Provider Network12-hour Update Trace (~180K

updates)

Page 10: FIB Aggregation

COMMENTS / QUESTIONS

Page 11: FIB Aggregation

ADDITIONAL SLIDES

Page 12: FIB Aggregation

FIB Aggregation: basic idea

04/20/23 12FIB Aggregation

/22

/23

A

Exploit aggregation opportunities over entire Table

Original Table Aggregated Table

A

/22

A

/23/23

AA

/22

ALevel 2SpecificsCombined(beyond L1)

Level 1SpecificsRemoved

Page 13: FIB Aggregation

Basic Idea for Updates

04/20/23 13FIB Aggregation

Example 2: Aggregate specifics [having same next hop] – Level2

/23/23

A

AA A

/22

With

dra

wWith

dra

w

Page 14: FIB Aggregation

Where does SMALTA stand?

04/20/23 14FIB Aggregation

AggregationOpportunities

Updates

Whiteholing

Level 1 Specifics removed Y N

Level2 Specifics combined Y N

Level 3 Specifics combinedover holes

Y Y

Level 4 Y Y

ORTC [1999]

Exploits all: Optimal N N

SMALTA Exploits all (~ORTC)

Y N

RIB snapshot Aggregate FIB: Snapshot Algo

BGP updates Aggregated table: Update Algo

Page 15: FIB Aggregation

Snapshot and WITHDRAW

04/20/23 15FIB Aggregation

Original Aggregated (SMALTA)C

B

A

A

C

A

A

A

BWithdraw

Withdraw

C

B A

C

A

B B

With Level 1-4Can’t aggregate

any further!

Deaggregation Opportunities to aggregate more

Page 16: FIB Aggregation

SMALTA Snapshot (300-400ms)~3-4x more processing than L1 and L2Applied infrequently

SMALTA Update~ same processing time as L1 and L2 (typical: 3s)Fewer avg. RIB-to-FIB downloads

Our view: another option for FIB aggregation

Remarks

04/20/23 FIB Aggregation 16

Page 17: FIB Aggregation

One-shot + ANNOUNCE + WITHDRAW

04/20/23 17FIB Aggregation

Original Aggregated(with SMALTA)

C

B

A

A

C

A

BAnnounce(

Q)Announce(

Q)

C

B

A

A

C

A

B QQ What if?

Withdraw

Page 18: FIB Aggregation

How far aggregated you are after N updates?

How long does it take to incorporate updates?

How many RIB to FIB downloads per update?

Incremental Updates: Analysis

04/20/23 18FIB Aggregation

Page 19: FIB Aggregation

Can’t aggregate entire table on every update Snapshot aggregation

Take current snapshot of RIB and Aggregate On “significant” routing changes (e.g., BGP hard reset) Perform a monolithic download after Snapshot

To reflect BGP updates in FIB Incremental updates to aggregated table

Practicalities

04/20/23 19FIB Aggregation