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

Report
FIB Aggregation
Zartash Uzmi
draft-uzmi-smalta-01
(with Ahsan Tariq and Paul Francis)
FIB Aggregation Work

First introduced: IETF 76



draft-zhang-fibaggregation
Level 1-4
SMALTA (at IETF78)

Better (near-optimal)
Normal Router Operation
RIB
Primary Routes
FIB
With Aggregation
Primary Routes
from RIB
Aggregate
FIB
11/7/2015
FIB Aggregation
2
Changes since IETF 78

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)
11/7/2015
FIB Aggregation
3
Evaluation of SMALTA

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




11/7/2015
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)
FIB Aggregation
4
L1/L2/SMALTA: Expectations?
Aggregated prefixes (as % of original)
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%
11/7/2015
For 2 Internet Gateway Routers (R1,R2)
and 5 Access Routers in Provider Network
FIB Aggregation
5
Aggregation and #next hops
% of Original
Fewer aggregation opportunities with more nexthops
Routeviews
12/2010
# Unique nexthops
Memory savings (for Tree Bitmap) are somewhat (~12%) lower
11/7/2015
FIB Aggregation
6
#Memory Accesses/Lookup time
% of Original
Lookup time (Tree Bitmap) varies in accordance with …
the #prefixes after aggregation
5 Access Routers
(Provider Network)
Effective # Unique nexthops
For Internet Gateway Routers, about 25% fewer
memory accesses when using Tree Bitmap
11/7/2015
FIB Aggregation
7
% of Original
% change in unaggregated
Incorporating Updates
Call to
Snapshot
function
An IGR (Internet
Gateway Router) in
Provider Network
12-hour Update
Trace
#updates (in 10K)
#aggregated prefixes is near-optimal after a
large number of updates are incorporated
11/7/2015
FIB Aggregation
8
Cumulative FIB downloads
FIB downloads per call to snapshot
Updates  FIB downloads
#Updates b/w consecutive
calls to “snapshot” function
#Updates b/w consecutive
calls to “snapshot” function
An IGR in Provider Network
12-hour Update Trace (~180K updates)
11/7/2015
FIB Aggregation
9
COMMENTS / QUESTIONS
ADDITIONAL SLIDES
FIB Aggregation: basic idea
Aggregated Table
Original Table
Level 1
Specifics
Removed
Level 2
Specifics
Combined
(beyond L1)
/22
/23
/23
A
A
/22
A
/22
A
A
A
/23
Exploit aggregation opportunities over entire Table
11/7/2015
FIB Aggregation
12
Basic Idea for Updates
Example 2: Aggregate specifics [having same next hop] – Level2
/22
/23
11/7/2015
A
A
/23
FIB Aggregation
A
A
13
Where does SMALTA stand?
Aggregation
Opportunities
Updates
Whiteholing
Level 1
Specifics removed
Y
N
Level2
Specifics combined
Y
N
Level 3
Y
Y
Level 4
Specifics combined
over holes
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
11/7/2015
FIB Aggregation
14
Snapshot and WITHDRAW
C
B
A
Original
Aggregated (SMALTA)
With Level 1-4 Deaggregation
Can’t aggregate  Opportunities to
any further!
aggregate more
C
A
A
Withdraw
A
Withdraw
A
B
C
C
A
B
A
B
11/7/2015
FIB Aggregation
B
15
Remarks

SMALTA Snapshot (300-400ms)
~3-4x more processing than L1 and L2
Applied infrequently

SMALTA Update
~ same processing time as L1 and L2 (typical: 3s)
Fewer avg. RIB-to-FIB downloads

Our view: another option for FIB aggregation
11/7/2015
FIB Aggregation
16
One-shot + ANNOUNCE + WITHDRAW
C
Original
Aggregated
(with SMALTA)
C
A
B
A
A
Announce(Q)
B
Announce(Q)
C
C
A
B
A
11/7/2015
A
Q
What if?
Withdraw
FIB Aggregation
B
Q
17
Incremental Updates: Analysis

How far aggregated you are after N updates?

How long does it take to incorporate updates?

How many RIB to FIB downloads per update?
11/7/2015
FIB Aggregation
18
Practicalities

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
11/7/2015
FIB Aggregation
19

similar documents