Download Globecom 09 Poster

WING
PreDA: Predicate Routing for DTN Architectures over MANET
The Web and
InterNetworking
Group
Flavio Esposito
[email protected]
Abstract
Ibrahim Matta
[email protected]
Predicate Routing in Action on Our UML Testbed
Data [MB]
e.g., direct all images captured by the camera on node S
!"#$
!!!"#$%&'()$!*+,)&-.!/"0
to DTNNetwork
node I (DTN)
for pre-processing
authorization
before
1.4
We consider a Delay Tolerant
whose usersand
(nodes)
are connected
by an underlying Mobile
34+
Summary of Contributions
sending
them
to
the
user
at
node
D.
!"#$%&'&(&)*&$+,-.&,&)/0/12)
1.2
Ad hoc Network (MANET) substrate. Users can declaratively express high-level policy constraints on how
From the network point of view, the predicate is a set of
“content” should be routed. For example, content can be directed through an intermediary DTN node 3+-4$#.$-'$!5(6$# 1
rules that each MANET packet has to satisfy. Our system
0.8
• We provide a reliable DTN neighbor discovery
for the purposes of preprocessing,
authentication,
etc.,
or
content
from
a
malicious
MANET
node
can
be
"<4
!"#
maps declarative user policies to such network-level routing
0.6
mechanism that leverages AODV’s HELLO mesdropped. To support such
content
routing
at
the
DTN
level,
we
implement
Predicate
Routing
[1]
where
)2:&
predicates. Predicates get propagated and installed as MANET0.4
!!!!!!!"#$%&'()$!*+,)&-.!1,22+#)!3+%$!!!
!
sages to propagate DTN node names. The converhigh-level constraints level
of DTN
nodes are
mapped
into low-level
predicates
the MANET
forwarding
rules.
Any DTN
node canrouting
inject, from
the within
0.2
93#="
$56&($7-0*&$
gence layer of the DTN stack then maintains the
application
level, aarchitecture
routing rule with
that gets
translated
nodes. Our testbed uses
a Linux system
Userseamlessly
Mode Linux
to emulate )2:&
every DTN node
0
0
100
200
300
400
500
mappings from DTN node names to IP (MANET)
intoImplementation
a MANET-levelcode.
rule,Inenabling
new
MANET prototype,
routing
Time [s]
with a DTN Reference
our initial
architecture
we use the On De8&()&.$92:;.&6
node addresses (Figure 1 and 2).
mand Distance Vector instructions.
(AODV) routing protocol at the MANET level. We use the network simulator ns-2
Figure 4: Data delivered at the destination vs.
We
next
show
two
examples
(Tables
I
and
II)
of
routing
(ns-emulation version) to simulate the wireless connectivity of both DTN and MANET nodes. Preliminary
• In addition to DTN node names, AODV’s HELLO
Fig.
1.
Architecture
of
a
DTN
node
running
over
a
MANET
substrate.
The
time
for
2%
packet
loss
probability.
predicates.
results show the efficient and correct operation of propagating routing predicates. For
theis application
of (PR) enabled. 2.5 2.5
node
Predicate Routing
messages are also used to propagate low-level
2.5
content re-routing through an intermediary, as a side effect, results
demonstrate
thepoint
performance
of
From
the network
of view, benefit
the predicate
is a set of
MANET routing predicates (Figure 3). These
!"#$
2
Drop Predicate
#1 Inserted
!!!"#$%&'()$!*+,)&-.!/"0
A.
Predicate
Examples
2
rules
that
each
MANET
packet
has
to
satisfy.
Our
system
Drop Predicate #1 Inserted
50
content re-routing that dynamically (on-demand) breaks the underlying end-to-end TCP connections into
34+ latter predicates are mapped by the convergence
2
Predicate
#1 Inserted
DropDrop
Predicate
#1 Inserted
maps declarative user policies to such network-level routing
!"#$%&'&(&)*&$+,-.&,&)/0/12)
B.
PreDA
Advantages
shorter-length TCP connections.
layer from given DTN-level requirements on routConsider the injection of two predicates
as
in
Table
I.
The
Drop
Predicate #1 Removed
45
Drop Predicate #1 Removed
predicates. Predicates get propagated and installed as MANET-1.5 1.5
Predicate
#1 Removed
DropDrop
Predicate
#1 Removed
1.5
3+-4$#.$-'$!5(6$#
objective of these two rules is to re-direct
the
traffic
destined
to
ing content.
level forwarding rules. Any Our
DTNPreDA
node can
inject, from
the cooperation between
architecture
enables
the
Drop Predicate #2 Inserted
Drop Predicate #2 Inserted
40
a node D to an intermediate DTN nodeapplication
I for authentication
or DTN
Drop Predicate #2 Inserted
level, a routing
rule that
gets
seamlessly
translated
1
overlay and the underlying network.
This cooperation
1
Drop Predicate #2 Inserted
"<4
• As !"#
a
proof
of
concept,
we
implemented
our
architec1
into
MANET-level
enabling
MANET
routing from the MANET level, Drop Predicate #2 Removed
pre-processing. The first rule directs all
dataa destined
to noderule,includes
(i) new
gathering
information
Drop Predicate #2 Removed
)2:&
35
ture
on our UML based testbed [3] that simulates a
Drop Predicate #2 Removed
instructions.
D but not yet pre-screened at node I to node I. Notice that if (ii) applying data mining techniques
!!!!!!!"#$%&'()$!*+,)&-.!1,22+#)!3+%$!!!
!
Drop
Predicate
#2
Removed
0.5
0.5 at the DTN level, and
DTN
Predicate Routing
0.5
network of emulated DTN-MANET nodes as well
Drop ifDrop
source
DTNis#4DTN #4
the API
destination node
93#="
if is
source
30the DTN overlay,
We the
nextintermediate
show two examples
(Tables I and
II) of routing
API D is in the path to reach
(iii) declaratively
generating
rules from
$56&($7-0*&$
Drop if0.08
source is
DTN 0.1
#6DTN #6
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Drop
source
Drop ifif0.09
source isis
DTN #4
)2:&
DTN Reference
Implementation
node I, then node D forwards the MANET
packets
matching
as MANET-only nodes. The wireless connectivity
Packet
Loss
Probability
predicates. Bundle Daemon (dtnd)
0
Drop if source
is DTN #6
thus facilitating general-purpose 0260network
applications.
For
270
280
290
300
310
320
330
260
270
280
290
300
310
320
330
Time (s)
0
Time
(s)
8&()&.$92:;.&6
Convergence
this ruleLayer
without reassembling the associated data message for example, a DTN network could use trust-based
260
270
280
290
300
and mobility of nodes are simulated using the ns
approaches
to 310 320 330
Time (s)
Bundle Router
Figure 5: Percentage improvement in data delivery
Application
LibraryExamples
(API)
pre-screening.
A.
Predicate
TCP
collect trust information and generate rules as in Table II, or a
simulator (ns-emulation version). The emulation
DTN
time
vs.
packet
loss
probability
with
95%
confidence
node
The
second
predicate
ensures
that
pre-screened
data
coming
sensed
signal
coming
from
a
node
can
trigger
the
generation
Predicate Routing Support Code
usesThe
UML (User Mode Linux) to run real DTN
TCP
UDP
... as in Table I. The2.5 Fig. 1. Architecture of a DTN node running over a MANET substrate.
Consider
the
injection
of
two
predicates
2.5
...
dtnsecrecv destinainterval
when
an authentication
predicate is injected.
from the intermediate DTN node I, dtnsend
reach the original
node
ispurpose
Predicate
Routing
(PR) enabled.
2.5
MANET
of
new
routing
predicates
(rules)
for
the
of
energy
Convergence
Layers
objective
of
these
two
rules
is
to
re-direct
traffic
destined
to
a
User Space
(reference implementation) and MANET (AODV
2.5 2.5
node
tion D. Note that if nodes (S, I and D)
areApplications
DTN
nodes,
then, saving,
2.5
load Ibalancing
or packetorprocessing.
2
Servernode
Libraries
Drop Predicate #1 Inserted
node
D
to
an
intermediate
DTN
for
authentication
routing) code.
Kernel Modules
2
Drop Predicate #1 Inserted
predicate overrides the normal DTN routing
process.
In
partic2
Drop Predicate #1 Inserted
pre-processing. The first rule directs all data destined to node 2 2
Predicate
#1 Inserted
DropDrop
Predicate
#1 Inserted
2
III.
O
VERALL
A
RCHITECTURE
ular, node I, not recognizing itself
as
the
destination,
would
Drop Predicate #1 Removed
• We
present throughput results showing the efficient
D 2:
butArchitecture
not yet pre-screened
at nodeimplementation.
I to node I. Notice that if1.5 and (iii) declaratively
Figure
of the DTN reference
generating
routing
rules
from
the
DTN
Drop Predicate #1 Removed
1.5
Drop Predicate #1 Removed
Figure 1: Architecture of ahave
PreDAdirected
enabled DTN
node running
received
bundles toTheDmodules
without
preprocessing.
1.5
the
destination
node
D
is
in
the
path
to
reach
the
intermediate
enabling PreDA are markedFigure
with a “1
star
”
.
and correct operation of propagating routing predoverlay,
thus
facilitating
general-purpose
network applications.
Drop
Predicate
#1 Removed
shows the DTN-MANET 1.5stack—our
modifiedDropDrop
and
1.5 1.5
Predicate
Removed
Predicate
#2#1
Inserted
over a MANET substrate. In
the stack
modified
and added supportsnode
I, then
node at
Dthe
forwards the MANET packets matching 1 For example, a DTNDropnetwork
Thus,
our our
PreDA
architecture
predicate
routing
could use trust-based approaches
Predicate #2 Inserted
added components are marked by 1“stars.”
Drop Predicate #2 Inserted
icates.
We
demonstrate
two
applications.
The
first
Drop
Predicate
#2
Inserted
this rule without reassembling the associated data message for to
components are marked by “DTN
stars.”level as well.
1
1
Drop Predicateand
#2 Inserted
collect
trust
information
generate
routing rules as in
1
Drop
Predicate
#2
Removed
The
block
named
Application
Programming
Interface
(API)
1
application
re-directs
content
to
an
intermediary
pre-screening.
Table
II,
or
a
sensed
signal
coming
from
a
node
can
trigger
Drop Predicate
#2 Removed
Drop Predicate
#2 Removed
0.5
Declarative approaches have been widely discussed as clean-slate alternative to routing or transport protocols
Our work
is
of [1,
the2].DTN
reference
implementation
is
extended
to
allow
Drop Predicate #2 Removed
0.5
node for pre-processing (Example 2, Figure 4
Drop Predicate #2 Removed
0.5
the
generation
of
new
routing
predicates
(rules)
for
the
purpose
Predicate
Action
0.5
The
second
predicate
ensures
that
pre-screened
data
comDrop
if
source
is
DTN
#4
0.5
the first to enable declarative routing into a DTN architecture overlayed over a MANET
substrate. to inject high-level requirements
applications
or
constraints.
0.5
Dropififsource
sourceisisDTN
DTN#4
#6
Drop
saving,
load balancing
src = ¬I ∧ dest= D ing
to I from the intermediate DTN node I, reach the original of energy
Drop ifor
sourcedata
is DTN #4processing.
and 5). We also demonstrate the correct operaDrop if source is DTN #4
0
0
Drop
if
source
is
DTN
#6
Drop
ifif source
is
#6
280
290
300310 310 320320
We
refer
to
this
modification
as
Predicate
Routing
API
(PRDrop
source
is DTN
DTN
#4#6
260
270 260
280270
290
300
330
Drop
if330
source
is DTN
src = I ∧ dest= D
to D
destination
D. Note that if nodes (S, I and D) are DTN0260 0260 2700 270 280 280 290 290Time300(s)Time
[s]
if source
DTN #6
tion of a second application where a malicious node
310 310 320 Drop
330 is
300
320
330
260
270
280
290
300
310
320
330
API).
Time (s)
TimeTime
(s)
nodes, then these predicates override the normal DTN routing 0260
III.
O
VERALL
A 320
RCHITECTURE
270
280
290
300 (s) 310
330
is isolated by dropping all its packets (Example
Time (s)
The
Predicate
Routing
Support
Code
(PRSC)
component,
TABLE I
Figure
6:
Throughput
drops
and
resumes
after
the
process. In particular, under normal DTN routing, node I,
D IRECT ALL D- TRAFFIC TO AN INTERMEDIATE
DTN NODE I itself
FOR
1,
Figure
6).
implemented
in
the
routing
protocol
user
space,
mainly
impleFigure
1
shows
the
DTN-MANET
stack—our
modified
not
recognizing
as
the
destination,
would
have
directed
“
dropping
”
predicates
propagate
and
are
removed.
A depicted PreDA packet together with the
DTNAUTHENTICATION
neighbor discovery
extension packet are attached with
CONTENT
.
andiptables
added Linux
components
received bundles to D without
preprocessing.
Thus, our
ments
two functionalities:
(1)PreDA
it uses the
facilityare marked by “stars.”
the HELLO message. Predicate info are disseminated together
with DTN
EID of newly discovered nodes,
architecture
supports
predicate
at the DTN
level routing
as
[9] torouting
install predicate
MANET
rules,
soblock
that MANET
The
named
Application
Programming
Interface
their MANET address, and their distance in hops to every other
well. node in the network.packets carrying content —IP packets
or DTN
(API)
of thebundle(s)—
DTN reference implementation is extended to
0
78
15 16
23 24
32
are routed based
constraints,
and to
(2)inject
it high-level routing requirements or
allow
applications
Predicate
Action on DTN-level routing
[1]
Timothy
Roscoe,
Steven
Hand,
Rebecca Isaacs, Richard Mortier, and Paul W. Jardetzky. Predicate Routing: Enabling Controlled
Predicate
Action
Predicate
constraints.
We referother
to this modification as Predicate Routing
Type = 15
Length
Flags
new routing extensions
to
discover
src = ¬I ∧creates
dest= Dand manages
to I
Type
Networking. Computer Communication Review, 33(1):65–70, 2003.
src
=
¬W
∧
dest=
D
drop
API (PR-API).
IP address
src = I ∧DTN
dest= nodes
D
to
D propagate MANET routing
and
predicates.
[2] Eiko Yoneki and Jon Crowcroft. Towards Data-Driven Declarative Networking in Delay Tolerant Networks. In DEBS 08 InterThe Convergence Layer (CL) interfacesThe
DTN
and MANET
Predicate
Routing Support
CodeSystems.
(PRSC)
component,
TABLE II
national
Conference
on
Distributed
Event-Based
ACM,
2008.
TABLEallI D traffic to an
Example
1:
Drop
traffic
not
origiExample
2:
Direct
Figure 3: Proposed PreDA D
packet
exby maintaining the mapping betweenimplemented
DTN node in
names
and protocol user space, mainly implethe routing
ROP TRAFFIC NOT ORIGINATED BY A White list W OF MANET NODES
D
IRECT ALL D- TRAFFIC TO AN INTERMEDIATE DTN NODE I FOR
nated
by a WhiteTOList
W of nodes
and
intermediate DTN node I for content au[3] Gabriele
Flavio Esposito,
and Ibrahim
Matta. Linux
Supporting
Predicate Routing in DTN over MANET. In CHANTS
tension, to propagate with DTN discovAND DIRECTED
A PRIVATE
NODE
.
ments
two
functionalities:
(1) it uses
the iptables
facility
CONTENT IP/MANET
AUTHENTICATION
.
addresses.
The mappings
areFerrari
usedAggradi,
to translate
directed to a node D.
thentication.
’08:[9]
Proceedings
of
the third MANET
ACM Workshop
onrules,
Challenged
Networks
at MOBICOM, pages 125–128, San Francisco, Caliery, piggybacked in HELLO messages.
to
install
predicate
routing
so
that
MANET
routing predicates on DTN node names to routing predicates
fornia,
USA, 2008.
ACM.content —IP packets or DTN bundles— are
packets
carrying
on corresponding IP/MANET node addresses.
routed based on DTN-level routing constraints, and (2) it
55
Overall PreDA Architecture
Data (MB)
(MB)
DataData
(MB)
Data (MB)
2
1.5
1
50
Percent improvement in delivery time
% Improvement in Delivery Time
2.5
45
40
35
4
4
4
4
0.5
3.8
30
3.8
3.8
3.8
0.01
3.6
0.02
3.6
0.03
0.04
3.6
if source0.1is DTN #4
0.08Drop0.09
0.05
0.06
0.07
Packet loss probability
Drop if source is DTN #6
3.6
0
260
3.4
270
280
3.4
290
3.4
300
Time (s)
310
320
330
3.4
3.2
3.2
3.2
3.2
3
3
3
3
2.8
2.8
2.8
2.8
2.6
2.4
2.4
2.4
2.2
2.4
2.2
2.2
2.6
DataData[MB]
(MB)
2.2
Data (MB)
(MB)
DataData
(MB)
Data (MB)
Data (MB)
2.5
2
2.6
2.6
2
2
2
0 0.2
2
0
0
1.5
0
0.2 0.4
0.2
0.4
0.4 0.6
0.2
0.4
0.6
0.6 0.8
0.6
0.8
0.8 1
1
1 1.2
1.2
1.21.4
1.4
1.4 1.6
1.6
1.61.8
1.8
1.8 2
2
0.8
1
1.2
1.4
1.6
1.8
2
2
1
4
4
4
0.5
4
4
3.8
3.8
4
3.8
3.8
3.6
Drop if source is DTN #6
3.8
0
260 3.6
3.6
270
3.4
Predicate Routing Design and Examples
Drop if source is DTN #4
3.8 3.6
280
290
3.6 3.4
300
Time (s)
310
320
330
3.6
3.4
3.4
3.2
3.4
3.2
3.4
3.2
3.2
3
3.2
3
2.8
3.2
3
3
3
2.8
2.6
3
2.8
2.6
2.8
2.8
2.4
2.8
2.6 2.2
References
2.6
2.6
2.4
2.6
2.4
2.2
2.2
2
2
0
2.4
2.4
2.4
2.2
2.2
2.2
0.2
2
0
0 0.2
2
0
2
2
0
0.2
0.4
0.4
0
0.2 0.4
0.2
0.2
0.6
0.6
0.4
0.4 0.6
0.4
0.8
0.8
0.6
0.6 0.8
0.6
1
1.2
1
1.4
1.6
1.8
1.6
2
1.2
1.4
1.8
2
0.8
0.8 1
1
1 1.2
1.2
1.21.4
1.4
1.4 1.6
1.6
1.61.8
1.8
1.8 2
2
0.8
1
1.2
1.4
1.6
1.8
2
2