Presentation 11/20: Network Layer Support for Service Discovery in MANETs
8 views
Skip to first unread message
Su Kim
Nov 20, 2007, 1:33:27 PM11/20/07
to ASU:CSE535 FALL 07 Mobile Computing
Authors: Ulas¸ C. Kozat and Leandros Tassiulas,
Title: "Network Layer Support for Service Discovery in MANETs,"
Published: INFOCOM2003, 2003
Title: "Network Layer Support for Service Discovery in MANETs,"
Published: INFOCOM2003, 2003
Aravind Krishna
Nov 27, 2007, 2:11:36 PM11/27/07
to asu_cse...@googlegroups.com
My critique is as follows.. :
Brief Summary:
The paper proposes a distributed service discovery approach that relies on the creation of a virtual backbone for registering and requesting services in a dynamic network topology. Virtual backbone is formed by a subset of nodes termed service broker nodes (SBNs). Each service provider element advertises its service information to the virtual backbone through its dominating node, and in this way the virtual backbone tries to maintain accurate information pertaining to the services provided by the network as a whole.
Pros:
Cons:
Thanks,
Aravind
Title: Network Layer Support for Service Discovery in Mobile Ad Hoc Networks
Brief Summary:
The paper proposes a distributed service discovery approach that relies on the creation of a virtual backbone for registering and requesting services in a dynamic network topology. Virtual backbone is formed by a subset of nodes termed service broker nodes (SBNs). Each service provider element advertises its service information to the virtual backbone through its dominating node, and in this way the virtual backbone tries to maintain accurate information pertaining to the services provided by the network as a whole.
Pros:
- They have pointed out the similarity between directory formation and maintenance, and the idea of clustering used in some proposed MANET network routing protocols.
- Protocols designed with this vertical integration approach are efficient at runtime.
- They assure that all nodes are part of the backbone or at least one hop away from it.
- The approach proves to be less costly than service discovery based on the multicast ODMRP protocol.
Cons:
- The scheme assumes that some routing protocols may exist underneath for supporting request/service delivery from clients to service providers or vise versa. This can make them inefficient for use in wireless ad hoc networks due to a redundant control packet flooding problem, possibly causing extremely serious network collision and latency of service acquisition.
- It is not clear how adaptable the virtual backbone is to node mobility; if backbone creation does not take mobility into account, the associated control traffic may saturate the network.
- The service discovery system does not provide mechanisms for service selection when multiple service instances of the same type coexist.
- Though run time performance is good, the design complexity is increased because, without a layered structure, low-layer protocols have to be developed separately for individual services.
- The creation and maintenance of multicast groups requires significant control message overhead.
Thanks,
Aravind
shayok.ch...@asu.edu
Nov 27, 2007, 2:13:05 PM11/27/07
to ASU:CSE535 FALL 07 Mobile Computing
Hi,
Please find my paper critique below:
Network Layer Support for Service Discovery in Mobile Ad Hoc
Networks
This paper presents a novel technique for service discovery in ad hoc
networks. The solution consists of two parts : a backbone management
phase (BBM) which produces a stable backbone network graph and a
distributed service discovery (DSD) phase which ensures successful
dissemination of the service discovery messages. The authors have
evaluated the proposed algorithm on three metrics, the mean control
message overhead, the hit ratio and the average time delay. Extensive
simulations show that the architecture supported by a virtual backbone
as described in the paper can perform well in a mobile environment.
Strong points:
1. The proposed technique is highly distributed which makes it robust
to link failures.
2. In the BBM phase, only a 1-hop local broadcast control message is
used for backbone formation.
3. The authors have given a mathematical proof of the time boundedness
and correctness of the BBM algo. This makes their approach more
convincing.
4. The authors have made a thorough analysis of their proposed
approach by changing a number of parameters (such as number of
clients, number of users, mobility level) and observing the results.
Weak points:
1. The approach has high communication overhead which renders it
unsuitable for large networks.
2. In the network model, it is assumed that all nodes have the same
transmission power. This may not be true in a practical situation.
3. In proving the correctness of the BBM algo, the authors have
assumed that the network graph is connected. This may not be true in
all cases.
Regards,
- Shayok.
> On Nov 20, 2007 10:33 AM, Su Kim <sujin...@gmail.com> wrote:
>
>
>
>
>
> > Authors: Ulas¸ C. Kozat and Leandros Tassiulas,
> > Title: "Network Layer Support for Service Discovery in MANETs,"
> > Published: INFOCOM2003, 2003
>
>
>
> Critique, Network Layer Support for Service Discovery in Mobile Ad Hoc Networks, paper.pdf
>
>
>
>
>
> > Authors: Ulas¸ C. Kozat and Leandros Tassiulas,
> > Title: "Network Layer Support for Service Discovery in MANETs,"
> > Published: INFOCOM2003, 2003
>
>
>
> 10KDownload- Hide quoted text -
>
> - Show quoted text -
swaroo...@asu.edu
Dec 4, 2007, 1:02:47 PM12/4/07
to ASU:CSE535 FALL 07 Mobile Computing
My Summary:
Summary: Network Layer Support for Service Discovery in
Mobile Ad Hoc Networks
In this paper, the authors have proposed a distributed service
discovery architecture for mobile ad hoc networks. This architecture
enables mobile nodes to discover services in an ever changing
environment and register to use those services.
Due to the infrastructure less nature of MANETs and mobility of the
mobile nodes, the nodes have little knowledge about other nodes, the
services they offer. Except for Bluetooth's SDP, no other service
discovery protocols are designed for wireless networks, although tons
of them exist for wired networks. Directory based service discovery
protocols place an overhead for MANETs, because the nodes on which the
directories reside often move out of the range of the other nodes.
Also, Energy consumption of directory nodes is more than those of
other nodes.
However, the authors show that the directory based service discovery
is a feasible solution for MANETs. Their solution involves two
phases,
Formation of a dynamic virtual backbone layer such that each node is
either a part of the backbone layer or one hop away from the backbone
layer
Broadcasting of service request and registration messages
The authors have presented detailed algorithms in both phases along
with several lemma and proofs for assumptions. The outline of their
algorithms are given below. The authors further claim that their
backbone selection phase is similar to the one used in VDBP, but the
way they incorporate the effect of link failures and interconnection
of VAP nodes are very different.
Backbone Management Phase: This algorithm can be described in three
phases
Initial Selection of Backbone nodes: Each node creates a neighborhood
information table and a routing table using the information carried by
the hello beacons.
Mesh Formation by finding the paths between backbone nodes: On the
basis of the developed routing tables, the nodes then build paths
amongst themselves. The path building operation is done in a finite
amount of time.
Maintenance against Topology changes: The maintenance of the
dominating set feature of the backbone is a very important task
against frequent topology changes.
Distributed Service Discovery Phase: Once the virtual backbone is
setup and the DA's are co located with the VAP nodes, the service
discovery phase begins.
When a server located on node i wants to register its service, it has
to register with the DA located on V APi assuming node i is a non-
backbone node.
V APi is referred as source VAP node. If the node i is already a black
node, then the
service is registered with the DA on the same node and node i itself
becomes the source VAP node.
Server may register with more DAs (even maybe with all DAs). Then we
need a multicast or broadcast mechanism to distribute the registration
messages to other DAs located on other VAP nodes.
Any time the VAP node of a server changes, it must renew its
registration with the DA operating on the new VAP node. Also, the
server should be able to keep the scope of its registration messages
local by bounding the number of black nodes the registration messages
could traverse.
When a client on node j requests for a service, node j forwards the
request to V APj provided node j is not already a black node and V APj
passes the request to colocated DA. If node j is black, then the
request is passed to the DA on node j. In case DAs do not have any
fresh registration for the service, the service should be requested
from other DAs again by multicasting or broadcasting.
Evaluation:
The authors performed the simulation of their proposed methodology on
the NS2 Simulator. The performance metrics they have considered are
total mean control message overhead
mean hit ratio
average time delay between a request and reply
On the basis of the evaluation, the authors came to the following
results
Under various mobility scenarios, number of users and servers, the
relative performances of the protocols remain same in general for a
fixed number of nodes and persistent service requests.
In terms of overhead, ODMRP is the most heavy-weight protocol and on-
demand anycast protocols are the most light-weight and an increase in
the number of users linearly affects the message overhead. DSDP is the
least sensitive protocol to increase in the number of users because
most of the overhead messages are transmitted periodically using hello
beacons
In terms of hit ratio anycast-AODV performs worst of all the protocols
while ODMRP performs the best, outperforming the second best protocol
DSDP by 18%.
Delay performance of DSDP is consistently better than other protocols.
This is because lower number of nodes are involved in search queries
and because registration of services makes the average distance in
number of hops shorter.
Positive points:
discovery protocols are vulnerable to buffer overflow attacks and
denial of service attacks. Best example of such a protocol is the
UPnP.
Conslusion:
In this paper the authors have presented an architectural solution for
distributed service discovery in MANETs. They have presented a two
phase algorithm for this purpose. They have further evaluated their
idea in comparison to various other protocols and user scenarios. The
results reveal that the directory architecture supported by a virtual
backbone system performs quite well in a MANET environment.
Summary: Network Layer Support for Service Discovery in
Mobile Ad Hoc Networks
In this paper, the authors have proposed a distributed service
discovery architecture for mobile ad hoc networks. This architecture
enables mobile nodes to discover services in an ever changing
environment and register to use those services.
Due to the infrastructure less nature of MANETs and mobility of the
mobile nodes, the nodes have little knowledge about other nodes, the
services they offer. Except for Bluetooth's SDP, no other service
discovery protocols are designed for wireless networks, although tons
of them exist for wired networks. Directory based service discovery
protocols place an overhead for MANETs, because the nodes on which the
directories reside often move out of the range of the other nodes.
Also, Energy consumption of directory nodes is more than those of
other nodes.
However, the authors show that the directory based service discovery
is a feasible solution for MANETs. Their solution involves two
phases,
Formation of a dynamic virtual backbone layer such that each node is
either a part of the backbone layer or one hop away from the backbone
layer
Broadcasting of service request and registration messages
The authors have presented detailed algorithms in both phases along
with several lemma and proofs for assumptions. The outline of their
algorithms are given below. The authors further claim that their
backbone selection phase is similar to the one used in VDBP, but the
way they incorporate the effect of link failures and interconnection
of VAP nodes are very different.
Backbone Management Phase: This algorithm can be described in three
phases
Initial Selection of Backbone nodes: Each node creates a neighborhood
information table and a routing table using the information carried by
the hello beacons.
Mesh Formation by finding the paths between backbone nodes: On the
basis of the developed routing tables, the nodes then build paths
amongst themselves. The path building operation is done in a finite
amount of time.
Maintenance against Topology changes: The maintenance of the
dominating set feature of the backbone is a very important task
against frequent topology changes.
Distributed Service Discovery Phase: Once the virtual backbone is
setup and the DA's are co located with the VAP nodes, the service
discovery phase begins.
When a server located on node i wants to register its service, it has
to register with the DA located on V APi assuming node i is a non-
backbone node.
V APi is referred as source VAP node. If the node i is already a black
node, then the
service is registered with the DA on the same node and node i itself
becomes the source VAP node.
Server may register with more DAs (even maybe with all DAs). Then we
need a multicast or broadcast mechanism to distribute the registration
messages to other DAs located on other VAP nodes.
Any time the VAP node of a server changes, it must renew its
registration with the DA operating on the new VAP node. Also, the
server should be able to keep the scope of its registration messages
local by bounding the number of black nodes the registration messages
could traverse.
When a client on node j requests for a service, node j forwards the
request to V APj provided node j is not already a black node and V APj
passes the request to colocated DA. If node j is black, then the
request is passed to the DA on node j. In case DAs do not have any
fresh registration for the service, the service should be requested
from other DAs again by multicasting or broadcasting.
Evaluation:
The authors performed the simulation of their proposed methodology on
the NS2 Simulator. The performance metrics they have considered are
total mean control message overhead
mean hit ratio
average time delay between a request and reply
On the basis of the evaluation, the authors came to the following
results
Under various mobility scenarios, number of users and servers, the
relative performances of the protocols remain same in general for a
fixed number of nodes and persistent service requests.
In terms of overhead, ODMRP is the most heavy-weight protocol and on-
demand anycast protocols are the most light-weight and an increase in
the number of users linearly affects the message overhead. DSDP is the
least sensitive protocol to increase in the number of users because
most of the overhead messages are transmitted periodically using hello
beacons
In terms of hit ratio anycast-AODV performs worst of all the protocols
while ODMRP performs the best, outperforming the second best protocol
DSDP by 18%.
Delay performance of DSDP is consistently better than other protocols.
This is because lower number of nodes are involved in search queries
and because registration of services makes the average distance in
number of hops shorter.
Positive points:
The proposed technique is highly distributed which makes it robust
to link failures.
to link failures.
The authors have made a thorough analysis of their proposed
approach by changing a number of parameters (such as number of
clients, number of users, mobility level) and observing the results.
approach by changing a number of parameters (such as number of
clients, number of users, mobility level) and observing the results.
They have pointed out the similarity between directory formation and
maintenance, and the idea of clustering used in some proposed MANET
network routing protocols.
maintenance, and the idea of clustering used in some proposed MANET
network routing protocols.
They assure that all nodes are part of the backbone or at least one
hop away from it.
Negative Points:
hop away from it.
The scheme assumes that some routing protocols may exist underneath
for supporting request/service delivery from clients to service
providers or vise versa. This can make them inefficient for use in
wireless ad hoc networks due to a redundant control packet flooding
problem, possibly causing extremely serious network collision and
latency of service acquisition.
It is not clear how adaptable the virtual backbone is to node
mobility; if backbone creation does not take mobility into account,
the associated control traffic may saturate the network.
The service discovery system does not provide mechanisms for service
selection when multiple service instances of the same type coexist.
Though run time performance is good, the design complexity is
increased because, without a layered structure, low-layer protocols
have to be developed separately for individual services.
In proving the correctness of the BBM algorithm, the authors have
for supporting request/service delivery from clients to service
providers or vise versa. This can make them inefficient for use in
wireless ad hoc networks due to a redundant control packet flooding
problem, possibly causing extremely serious network collision and
latency of service acquisition.
It is not clear how adaptable the virtual backbone is to node
mobility; if backbone creation does not take mobility into account,
the associated control traffic may saturate the network.
The service discovery system does not provide mechanisms for service
selection when multiple service instances of the same type coexist.
Though run time performance is good, the design complexity is
increased because, without a layered structure, low-layer protocols
have to be developed separately for individual services.
assumed that the network graph is connected. This may not be true in
all cases.
There are also security concerns about the protocol, as service
all cases.
discovery protocols are vulnerable to buffer overflow attacks and
denial of service attacks. Best example of such a protocol is the
UPnP.
Conslusion:
In this paper the authors have presented an architectural solution for
distributed service discovery in MANETs. They have presented a two
phase algorithm for this purpose. They have further evaluated their
idea in comparison to various other protocols and user scenarios. The
results reveal that the directory architecture supported by a virtual
backbone system performs quite well in a MANET environment.
Reply all
Reply to author
Forward
0 new messages