An Anycast-based P2P Routing Protocol for Mobile Ad Hoc Networks* Rong Cheng, Hai Jin, Ke Shi, Bin Cheng Cluster and Grid Computing Lab Huazhong University of Science & Technology, Wuhan, 430074, China [email protected]

Abstract With the increasing popularity of mobile ad-hoc networks (MANETs), it becomes a challenge to integrate peer-to-peer (P2P) technique into MANETs efficiently. This paper proposes to integrate anycast routing technique to mobile P2P networks to reduce overhead and optimize query lookup service. We propose a novel anycast-based mobile P2P routing protocol for MANETs, called AMPP, which integrates Chord and anycast routing protocol at the network layer in highly dynamic MANETs. Our simulation results show that integrating anycast routing technique to mobile P2P network can reduce query lookup cost and improve network performance.

1. Introduction Mobile ad hoc networks (MANETs) and peer-topeer (P2P) overlay networks are two popular research areas. Some highly structured P2P lookup algorithms have been proposed, such as Chord [1], Pastry [2], Tapestry [3], and CAN [4], which all employ Distributed Hash Table (DHT) technique. However, these P2P protocols only work in wired networks. P2P networks and MANET share some key characteristics: service location and discovery, selforganization and decentralization. Both need to solve the same fundamental issue, connectivity [5]. Unfortunately, P2P and MANET research communities have been working largely separately. With the increasing popularity of MANETs, it becomes necessary to efficiently integrate P2P technique into MANETs. However, bandwidth limitations, node mobility, and multi-access interference pose unique challenges to efficiently deploy P2P applications in MANETs. There are two approaches to implement structured P2P protocol in .

* This paper is supported by National Science Foundation under grants No.60433040.

MANETs: a layered approach which directly overlays P2P protocol on top of an existing multi-hop routing protocol for MANETs, and an integrated approach which integrates a P2P protocol with a multi-hop routing protocol at the network layer in MANETs [6]. However, the layered approach induces multiple layer redundancy and duplication in terms of messages and communication between nodes [7]. In [8], an integrated approach which integrates Pastry [2] and DSR [9] protocol at the network layer is proposed. However, this approach relies on flooding of packets to discover and maintain routes and is based on unicast routing. We propose to integrate anycast routing technique to mobile P2P networks to reduce routing overhead and optimize query lookup service. Anycast, as an elegant solution for service discovery in the Internet, is a popular IP communication model, just like unicast and multicast [10]. Anycast essentially provides a means to locate and communicate with any one of a set of distributed servers or service access points within a network. IP anycast can provide automatic service discovery [11]. Anycast is more important for an ad-hoc network in terms of resource, robustness and efficiency [12]. Anycast technology and related dynamic routing functionality provides significant improvements to mobile network architectures [13]. In

MANETs, anycast routing protocol can balance the network load and reduce the delay of packet and improve the network throughput. Anycast service can also improve performance of ad hoc network when mobility and link disconnection are frequent [12]. There are three styles of IP anycast usage, service discovery, query/reply services, and routing services. They can be used for easy configuration, and improving robustness and efficiency for many applications or lower-layer protocols [14]. We deploy IP anycast in DHT-based Chord protocol at network layer in MANETs, both for new DHT members to find existing ones (service discovery), and for nonmembers to route queries to the DHT (routing service).

In this paper, we propose an anycast-based mobile P2P routing protocol for MANETs, called AMPP, which integrates Chord and anycast routing protocol at the network layer in MANETs. Our simulation results show that AMPP protocol can reduce overhead of network and improve network performance. The paper is structured as follows. In section 2, we give some related works. We propose AMPP protocol in section 3, and perform simulation experiments in section 4. Finally, we conclude the paper and point out some future works related to this topic.

2. Related Works Several studies have been done for P2P system in MANETs. The first approach to establish P2P file sharing for MANET is the protocol 7DS [15]. 7DS uses local broadcast transmissions for sharing web documents among peers in order to enable online web browsing without connected to the Internet. DPSR MANET routing protocol is suggested in [16], which applies Pastry to a MANET routing protocol DSR. A routing protocol named Ekta is proposed in [7]. DPSR and Ekta share the essence of integrating Pastry and DSR at the network layer. This is because that Pastry can find requested information more efficiently (namely O(logn)) than the regular MANET on-demand routing protocols. Both DPSR and Ekta use broadcast technique, resulting in flooding the whole network whenever a search for a node requires. RINGS is proposed to reduce flooding at a certain extent [6]. However, RINGS does not support anycast. Proem is a P2P Platform for developing mobile P2P applications [17]. Proem Mobile supports 802.11b in ad hoc mode. But Proem does not consider the multihop mobile ad hoc networks. A cross-layer protocol that embeds a search mechanism in a geographical location service is proposed [18]. Since it is first introduced in RFC 1546 [9], IP anycast has been a popular technology. However, current researches in anycast technology mostly focus on wired network. For example, IP anycast is used for root Domain Name Service (DNS) servers in wired network. Anycast has potential implications on improving performance and providing easier access to distributed services and also has implications on improving the robustness of ad-hoc networks, especially given their highly dynamic nature. In general, IP anycast can be used as a simple, scalable, and robust service discovery mechanism. P2P technology can be used as service location mechanism in distribution mobile environment. Our AMPP protocol deploys IP anycast in DHT-based P2P application in MANETs.

3. AMPP Protocol Design AMPP integrate DHT protocol operating in a logical namespace and MANET routing protocol operating in a physical namespace. The key idea of the integration is to bring the Chord protocol to the network layer of MANETs via a one-to-one mapping between the service keys of the mobile nodes providing services and their anycast-groupIds in the namespace. In AMPP, Chord provides support for just one operation: given a service key, it maps the service key onto an anycast group providing certain service. Chord is used to locate the group providing a service and anycast routing technology is used to choose the best anycast member from the group. In a Chord network, nodeId is a unique identifier of a node in the overlay network. The mapping between nodeId and IP address is assumed to be one-to-one. In AMPP, we extend the above assumption to allow multiple nodes to share the same anycast groupId to incorporate anycast. These nodes form an anycast group. AMPP assigns unique 32-bit anycast-groupId to an anycast group in a MANET by hashing the IP address of a mobile node of the same anycast group by using a collision-resistant hashing function such as SAH-1 [20]. All members of each anycast group share the same anycast-groupId. In AMPP, service keys are hashed from services identifiers. The format of the Route Request (RREQ) packet in AMPP routing protocol is illustrated in Figure.1. In order to support Chord in network layer, we extend the original format of the Route Request (RREQ) packet of AODV protocol [19] with two new fields (Service key and anycast-groupId). Service keys are hashed from services identifier (or files identifier), and are received from application layer. Anycast-groupId is hashed from the IP address of certain anycast member node of the same anycast group.

Fig.1. Format of Route Request (RREQ) packet The format of the Route Reply (RREP) packet in AMPP routing protocol is illustrated in Figure.2. We add one new field (anycast-groupId) in the original

AODV RREP packet format. The format of other packet, such as Route Error (RRER) packet, need not be changed.

packet with service key and anycast-groupId s to other neighbors. Figure 3 shows an example of service location and discovery using our AMPP protocol. In Fig.3, (ki, G(Si)) denotes that anycast group Si provides service ki, and RREQ(ki, Si) denotes a Route Request (RREQ) packet with service ki and anycast-groupId Si. Node D, E and F are in group G(S1) providing service k1, and node J, K and H in group G(St) providing service kt.

Fig. 2. Format of Route Reply (RREP) packet In AMPP, each node maintains a routing table. The routing table entry contains anycast-groupId, Destination IP Address, Destination sequence Number, Interface, Hop Count, Last Hop Count, Next Hop, List of precursors, Lifetime, Routing Flags. AMPP protocol is based on AODV protocol. The difference lies in the content of each routing table entry. One new field (anycast-groupId) is added to the destination in each routing table entry in AMPP protocol. Whenever a source node wants a particular service (or data), a service key is first generated by hashing the service identifier in application layer and then is sent to network layer. If the source node is a unicast node, it first routes queries to the DHT and DHT members (anycast members) locate the requested service, and by using IP anycast it can discover the closest service provider. If it is a DHT member (anycast node), it locates the requested service using Chord’s lookup algorithm. Following we consider how to process service queries for an intermediate node. (i) Current node n is a unicast node Because a unicast node does not provide any service, it broadcasts a Route Request packet (RREQ) with service key and a default anycast-groupId just for unicast node to neighbors. (ii) Current node n is a DHT member (an anycast node) If node n provides service key and is not the source node, a Route Reply (RREP) packet containing destination IP address and corresponding anycastgroupId is sent back to the source node. Otherwise it checks this group’s finger table and selects an anycastgroupId s closest to the key using Chord’s lookup algorithm. Then it checks all the routing table entry whose anycast group Id equals to s to find whether there exist a route for this communication. If there are many routes, it will choose the route with minimum hop count and the corresponding destination as server node. If there does not exist a route to any node in group s, it will re-broadcast the Route Request (RREQ)

Fig. 3. An example of service location and discovery When a unicast node A needs service kt, which is generated by hashing this service identifier, it broadcasts a Route Request (RREQ) packet with kt and a constant DEFAULT just for unicast node to neighbors. When a unicast node B receives RREQ(kt, DEFAULT) packet from node A, it re-broadcasts RREQ(kt, DEFAULT) packet. When an anycast node D in group G(S1) receives a RREQ packet, it locates service kt in the closest group G(S2) using Chord’s lookup algorithm. Then it broadcasts RREQ(kt, S2) packet. When a RREQ(kt,St) packet arrives at group G(St) providing service kt, anycast node H sends a Route Reply (RREP) packet with the information of destination IP address and corresponding anycastgroupId to the source node A. Node A checks all the entry of its routing table and selects the closest route for anycast-groupId equals to St. We assume that the number of anycast groups is steady in a MANET, so we consider the instance that a node joins or leaves an anycast group. When a mobile node attains a service, such as a file and a particular data, it can share its resource to other nodes and becomes a mobile peer. Whenever a mobile node joins or leaves an anycast group, it originates a request packet and sends it to network layer. The anycastgroupId is updated correspondingly, and broadcasts the information about updating information.

with the increasing of node mobility.

14

x 10

5

2.5

2

1.5

1

0.5

0

0

100

200

300

400

500

Pause time (seconds)

Fig. 5. Routing overhead for 20 traffic sources 5

x 10

5

Layered AMPP

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0

4

50

100

150

200

250

300

350

400

450

500

Pause time (seconds)

Fig. 6. Routing overhead for 30 traffic sources

Layered AMPP

12 Routing overhead (packets)

x 10

Layered AMPP

Routing overhead (packets)

We use simulation model ns-2 to evaluate the performance of AMPP protocol and the layered approach with Chord directly layered on top of MANETs. In the layered approach, Chord is implemented over AODV as an application in ns-2. AMPP, being a network-layer protocol like AODV, is implemented as a routing agent in ns-2. The random waypoint model [21] is used in which 50 nodes move at a speed uniformly distributed between 1-19m/s in an area of 670m×670m. During the 500 seconds long simulation process, all the data packets, each with the size of 512-byte, in the network are sent by the continuous bit rate (CBR) with initiating packets at the rate of 1 packets/second. There are four anycast groups and each anycast group has five anycast members. We control the offload of whole network by varying mobility and the number of traffic sources. We evaluate the protocol performance by counting the routing overhead, the number of control packets at network layer. There are 6 different mobility models in our tests on mobile ad hoc network of 50 nodes. They use 100k (k=0, 1, 2, 3, 4, 5) seconds as node mobility pause time, which denote different node mobility. Figure 4 shows the routing overhead of AMPP and the layered approach as the network mobility is varied in case of 10 traffic sources. It can be found that the routing overhead of AMPP is much lower than that of the layered approach for all mobility models.

3

Routing overhead (packets)

4. Performance Evaluation

From the comparison above, it is evident that AMPP protocols can reduce overhead of network layer and improve network performance.

10 8

5. Conclusion and Future Work

6 4 2 0

0

100

200

300

400

500

Pause time (seconds)

Fig. 4. Routing overhead for 10 traffic sources To evaluate the overhead generated by both approaches, we also vary the number of traffic sources. From Figure 5 and Figure 6, it can be found that as the traffic sources increases to 20, the difference between AMPP and the layered approach is more obvious but the overhead of two approaches are both descending

In this paper, we propose a network layer routing protocol AMPP, which can reduce overhead and optimize query lookup service. AMPP combines advantages of IP anycast and DHT-based P2P protocol. Our simulation results show the improvement from our initial design of AMPP. Our further works include the optimization of AMPP and integration of a P2P file sharing application as a practical example of using AMPP.

References [1]

I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan, “Chord: A Scalable Peer-to-peer

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

Lookup Service for Internet Applications”, Proc. ACM SIGCOMM’01. A. Rowstron and P. Druschel, “Pastry: Scalable, distributed object location and routing for large-scale peer-to-peer systems”, Proc. of Middleware, November 2001. B. Y. Zhao, J. D. Kubiatowicz, and A. D. Joseph, “Tapestry: An Infra- structure for Fault-Resilient Wide-area Location and Routing”, Technical Report UCB//CSD-01-1141, U. C. Berkeley, April 2001. H. Balakrishnan, F. Kaashoek, D. Karger, R. Morris and I. Stoica, “Looking up Data in P2P Systems”, Communications of the ACM, 46(2), Feb. 2003, pp.43-48. L. Yan, K. Sere, X. Zhou, and J. Pang, “Towards an Integrated Architecture for Peer-to-Peer and Ad Hoc Overlay Network Applications”, Proc. of 10th IEEE International Workshop on Future Trends in Distributed Computing Systems (FTDCS’04), May 26 - 28,2004, Suzhou, China. H. Pucha, S. M. Das, and Y. C. Hu, “How to Implement DHTs in Mobile Ad Hoc Networks”, Proc. of 10th ACM International Conference on Mobile Computing and Network (MobiCom’04), Philadelphia, PA, September 26-October 1, 2004. K. Patel, S. Iyer, and K. Paul, “RINGS: Lookup Service for Peer-to-Peer Systems in Mobile Ad-hoc Networks”, Proc. of 6th International Workshop on Distributed Computing (IWDC’04), Kolkata, India, December 27-30, 2004. H. Pucha, S. M. Das, and Y. C. Hu, “Ekta: An Efficient DHT Substrate for Distributed Applications in Mobile Ad Hoc Networks”, Proceedings of the 6th IEEE Workshop on Mo- bile Computing Systems and Applications (WMCSA’04), English Lake District, UK, December 2-3, 2004. D. B. Johnson and D. A. Maltz, Dynamic Source Routing in Ad Hoc Wireless Networks, Kluwer Academic Publishers, 1996. J. Hagino and K. Ettikan, “An Analysis of IPv6 Anycast”, draft-ietf-ipngwg-ipv6-anycast-analysis02.txt, Internet Draft, IETF, June 2003. C. Partridge, T. Mendez, W. Milliken, “Host Anycast

[12]

[13] [14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

Service”, RFC 1546, 1993. J. Wang, Y. Zheng, and W. Jia, “An AODV-Based Anycast Protocol in Mobile Ad Hoc Network,” Proceedings of 14th IEEE International Symposium on Personal Indoor and Mobile Radio Communication, 2003. V. Park and J. Macker, “Anycast Routing for Mobile Networking”, Proc. IEEE MILCOM 99, 1999 H. Ballani and P. Francis, “Towards a Global IP Anycast Service”, Proc. of SIGCOMM 2005, August, 2005. M. Papadopouli and H. Schulzrinne, “Effects of power conservation, wireless coverage and cooperation on data dissemination among mobile devices”, Proceedings of ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), pp.117–127, October 2001. Y. C. Hu, S. M. Das, and Hi. Pucha, “Exploiting the Synergy between Peer-to-Peer and Mobile Ad Hoc Networks”, Proceedings of 9th Workshop on Hot Topics in Operating Systems (HotOS-IX), Lihue, Kauai, Hawaii, May 18-21, 2003. G. Kortuem, J. Schneider, D. Preuitt, T. GC Thompson, S. Fickas, and Z. Segall, “When Peer-toPeer Comes Face-to-Face: Collaborative Peer-to-Peer Computing in Mobile Ad hoc Networks”, Proc. of Inter. Conference on Peer-to-Peer Computing, Linköpings, Sweden, August 2001. G. Ding and B. Bhargava, “Peer-to-peer file-sharing over mobile ad hoc networks”, Proceedings of IEEE Annual Conference on Pervasive Computing and Communications Workshops, pp.104-109, March 2004. C. E. Perkins, E. M. Belding-Royer, and Samir R, Das, “Ad hoc On-demand Distance Vector (AODV) Routing”, Internet draft, IETF, 2001. FIPS 180-1, Secure Hash Standard, U.S. Department of Commerce/- NIST, National Technical Information Service, Springfield, VA, Apr. 1995. J. Yoon, M. Liu, and B. Noble, “Random waypoint considered harmful”, Proc. of IEEE INFOCOM, April 2003.