Investigation of Global Network Routing Behavior

1
Investigation of GlobalNetwork Routing Behavior

• BJ Premore
• Dartmouth College
• Prof. David Nicol, Advisor
• October 30, 2000
• In collaboration with
• Jim Cowie, Renesys Corporation
• Tim Griffin, ATT Labs-Research
• Andy Ogielski, Renesys Corporation
• and several other colleagues

2
Overview

• Objectives
• To better understand the dynamics of BGP
  behaviors (route fluctuations, failure modes,
  convergence)
• Provide the ability to explore the global impact
  of implementation tradeoffs, as well as
  non-standard extensions/modifications to the
  protocol
• Combined use of the black box (Internet
  measurements) approach and the glass box approach
  (simulations)
• Implementation and Functionality
• SSFNet BGP-4 architecture
• Finite state machine, peering session
  establishment, timing, iBGP, route reflection,
  messages and path attributes, decision process
  route selection, policy-based filtering, feedback
• Validation
• Research Applications
• Timing (Min Route Adver), route flapping (timing
  again), policy interaction convergence, much
  much more

3
loose confederation of independent networks (ASes)
Global Internet
the only glue that holds it all together
BGP
4

• most of the conventional wisdom about
• routing convergence is not accurate

Craig Labovitz, Leiden 2000
5
Interior vs. Exterior Routing Protocols
Interior Gateway Protocols (IGP) inside
autonomous systems
Exterior Gateway Protocols (EGP) between
autonomous systems
AS 701
UUNet
OSPF, IS-IS, RIP, EIGRP, ...
BGP
Metric Based
Policy Based
ATT WorldNet
Bell Labs
AS 144
AS 7018
Tim Griffin, Leiden 2000
6
AS graphs obscure topology!
The AS graph may look like this.
Tim Griffin, Leiden 2000
7
What to do?
black box
glass box
transparent and complete simulation model of
BGP routing on a global scale
multi-router monitoring of BGP routing tables and
route updates
These approaches are complementary
8
Pros and Cons

• We cant
• expect to model real-world routers perfectly with
  every detail
• We can
• capture the most important characteristics
• change and tweak the protocol
• explore consequences of fundamental design of BGP
• explore proposed and novel protocol extensions
• evaluate and analyze collective behavior on a
  large-scale

9
SSF, Raceway and SSFNet

• SSF
• Modern software for modeling and simulation of
  very large networks
• Raceway
• A fast Java implementation of SSF
• SSFNet
• A collection of packages for building IP network
  models
• Hosts, routers, networks, major Internet
  protocols
• Compositional approach to large network design
  with CIDR-compliant automatic IP address
  allocation
• Includes our BGP-4 implementation

10
SSF.OS.BGP4

• Based on RFCs
• RFC 1771 BGP-4 and latest drafts
• RFC compliant implementation
• Includes some RFC-specified extensions (Route
  Reflection)
• Has features similar to those used by vendors
  (policy-based filtering)

11
SSF.OS.BGP4 Functionality

• Finite state machine, timers, RIB
• TCP transport
• Peering exterior and interior
• Route reflection
• Messages and path attributes
• Policy
• Filter based on path attribute
• Attribute modification
• Monitoring of protocol operation

12
Package SSF.OS.BGP4 Organization
BGPSession
PeerEntry
PeerEntry
RIBIn
LocRIB
RIBOut
RIBOut
RIBIn
Policy Rule (inbound)
Timers ConnRetry KeepAlive Hold MinAdver
Policy Rule (outbound)
13
Validation Methodology

• No standards, create our own suite
• Basic behavior in simple topologies
• Peering session maintenance (Hold KeepAlive
  timer operation)
• Route advertisement and withdrawal
• Route selection
• Reflection
• Internal BGP
• General behavior in complex topologies
• End-to-end data delivery
• Exercises basic behaviors as well
• Policy testing
• Converging and non-converging gadgets Griffin
  1999

14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Interesting Possibilities

• Sender-side loop detection Labovitz 1999
• Better value for MinAdver timer?
• Improved route flap dampening?
• Policy studies
• How do various configurations affect convergence?
• Test effects of policy changes before deployment
• EGP-IGP interaction studies
• Are there instability side-effects?
• Is it safe to convert between different cost
  metrics?

18
CDF of BGP Routing Table Convergence Times
Craig Labovitz, Leiden 2000
New Route Long-gtShort Fail-over
Short-gtLong Fail-Over
Failure

• Less than half of Tdown events converge within
  two minutes
• Tup/Tshort and Tdown/Tlong form equivalence
  classes
• Long tailed distribution (up to 15 minutes)

19
End-to-End Impact of Failover
Craig Labovitz, Leiden 2000

• ICMP loss to 100 randomly chosen web sites with
  VIF source address of our probe
• Tlong/Tshort exhibit similar relationship as
  before

20
An Experiment with SSF.OS.BGP4

• The Model
• Topology full mesh of N ASes, each with just 1
  router
• No route filtering
• Shortest path is best
• Advertise, Withdraw, Wait and Watch
• Wait for system to reach stable state, then
• AS 1 advertises a bogus destination to everyone
  else
• Wait for system to reach a stable state again,
  then
• AS 1 tells everyone that the bogus route is not
  reachable through it any more
• Wait for system to reach a stable state again

21
4
5
1
bogus
3
2
N 10 20 30 40 50
longest path 9 20 28 40 46
convergence time after withdrawal (sec) 150
480 720 1080 1260
avg updates due to withdrawal (range) 59.50
(35-84) 269.55 (58-397) 539.10 (118-892)
945.20 (160-1647) 1423.66 (196-2377)
22
. . . 1610.040778415 bgp_at_381 snd update to
bgp_at_21 wdsbogus 1610.040778415 bgp_at_381 snd
update to bgp_at_201 wdsbogus 1610.040778415
bgp_at_381 snd update to bgp_at_321
wdsbogus 1610.040778415 bgp_at_381 snd update
to bgp_at_441 wdsbogus 1610.040890567 bgp_at_321
snd update to bgp_at_381 nlribogus,asp32 44 34 38
4 22 2 20 48 10 26 12 6 16 36 8 14 24 28 41 18 51
21 33 45 43 35 3 5 47 23 31 37 49 25 46 39 7 27
13 9 29 11 15 17 50 19 42 40 30 1 1610.040890567
bgp_at_321 snd update to bgp_at_441
wdsbogus 1610.040907352 bgp_at_441 snd update
to bgp_at_381 wdsbogus 1610.040907352 bgp_at_441
snd update to bgp_at_341 nlribogus,asp44 38 34 32
4 22 2 20 48 10 26 12 6 16 36 8 14 24 28 41 18 51
21 33 45 43 35 3 5 47 23 31 37 49 25 46 39 7 27
13 9 29 11 15 17 50 19 42 40 30 1 1610.050930294
bgp_at_441 snd update to bgp_at_321 wdsbogus . . .
23
Future Work

• Add features
• Route flap dampening
• Communities
• Etc.
• Do experiments
• Look for better timer values
• How policy affects convergence
• Improve route flap dampening
• Test extensions and other proposed modifications
• Much much more

24
Information

• SSF/Raceway and SSFNet
• http//www.ssfnet.org/
• SSF.OS.BGP4
• http//www.cs.dartmouth.edu/beej/research/bgp/jav
  a/
• (or follow link from ssfnet.org)

25
SSF.OS.BGP4 Primary Classes

• BGP session
• An instance of BGP on a router (BGP speaker)
• Primary BGP engine configuration, FSM, timing,
  decision process, updates
• Contains peer entries, RIB (In, Out, Local), Min
  AS Origination Timer
• PeerEntry
• Represents a peer ( associated peering session)
• Contains filters (incoming and outgoing) and
  timers (ConnRetry, KeepAlive, Hold, MinAdver)
• AdjRIBIn, LocRIB, AdjRIBOut
• MinAdverTimer, EventTimer
• Policy.Rule

26
Coming Soon

• Aggregation
• Route flap dampening
• Several extensions
• Communities
• Confederations
• Etc.

27
This sample DML code configures an AS with a
single router running BGP. It performs explicit
configuration of all BGP attributes. It is taken
from the goodgadget validation test.
(continued next page)
Net id 1 AS_status boundary
router id 1 graph
ProtocolSession name bgp use
SSF.OS.BGP4.BGPSession autoconfig
false connretry_time 120
min_as_orig_time 15 reflector false
neighbor as 0 address 1(1)
use_return_address 1(1) hold_time 90
keep_alive_time 30 min_adver_time 30
infilter give low priority to routes learned
from 0 clause
precedence 1 predicate
action primary
permit atom attribute
local_pref type set value 80

outfilter _extends .filters.permit_all

28
neighbor as 2 address
1(2) use_return_address 1(2)
hold_time 90 keep_alive_time 30 min_adver_time
30 infilter give high priority to
routes learned from 2 clause
precedence 1 predicate
action
primary permit atom
attribute local_pref type set value 100

outfilter _extends .filters.permit_all
neighbor as
3 address 1(2) use_return_address 1(3)
hold_time 90 keep_alive_time 30
min_adver_time 30 infilter deny
all routes learned from 3 clause
precedence 1 predicate action primary deny
outfilter
_extends .filters.permit_all
ProtocolSession name socket use
SSF.OS.Socket.socketMaster
ProtocolSession name tcp use
SSF.OS.TCP.tcpSessionMaster
ProtocolSession name ip use SSF.OS.IP
interface idrange from 0 to 3
host id 101 _extends .basic_host
link attach 1(0) attach 101(0) delay 0.001