Computer Communications 28 (2005) 1887–1902 www.elsevier.com/locate/comcom

SNMP information based routing mechanism for fast handoff in mobile IP Sang-Hoon Ryu* Computer Science, Korea University, Software System Laboratory, Department of Com., 5ga Anamdong Seongbukgu, Seoul 136701, South Korea Received 16 January 2004; revised 23 February 2005; accepted 9 March 2005 Available online 26 April 2005

1. Introduction Mobile IP is standard protocol to support mobility for current internet systems. In mobile IP network, basically, there is a router called home agent (HA) at the home network of a mobile host (MH) in which a MH registers its own care of address (CoA) that is used as address by a MH at external network. After that, HA sends its packets binding for a MH to HA having CoA of the destination MH [1]. Accordingly, whenever the HA takes handoff in the mobile network, the node shall register its own IP address, namely CoA, into HA. Mobile IP suffers from triangle routing problem. That is, hereafter the packets sailing for the mobile node would not be routed directly for the destination, but always make a detour by way of HA to the target node. This kind of issue is called triangle routing problem. Particularly, the IPv6 of next version of IPv4 adopting ‘binding cache’ as a new routing optimization method for triangle routing problems can decrease network load and packet propagation delay as embedding the ‘binding cache’ into a piece of protocol. Nevertheless, hierarchical MIPv6 reduces round trip time (RTT) and signaling regionally by introducing a new function, the mobility anchor point (MAP). This means in order to make correspond node (CN) and HA aware of changing of routing path to a MH after a MH moves another coverage of MAP, a MH sends binding update (BU) message to CN and HA. In case global handoffs occur continuously, this makes multimedia application device have more heavy burden because it requires to add one more RTT and signaling to legacy handoff delay between CN and MH and HA [2]. * Tel.: C821039254569; fax: C8229219137. E-mail address: [email protected].

0140-3664/$ - see front matter q 2005 Published by Elsevier B.V. doi:10.1016/j.comcom.2005.03.002

Just because mobile IP is suggested for the purpose of supporting mobility of node, it dose not support simple new connection to the network node, but guarantee quality of service (QoS) after the mobile node hands off. In mobile IP networks, specific QoS traits such as throughput, delay time and error rate are matters of great important factors in case of supporting real time service or multimedia streaming service at the mobile environments [3]. In particular, the core requirements of real time service have something to do with packet delay, especially, have a close relation to hand off delay. For that reason, accordingly, if circumstances admit to ensure a high level QoS at the mobile network environments, we have had lively discussion on the problems of hand off delay as a new issue. In this paper, to make up for these shortcomings, we begin by simple network management network (SNMP) information based hierarchical routing which has been derived from information based routing in active network adding Keyup procedure proposing keyword management methodology. By this new routing mechanism, we argue for fast handoff and elevating QoS to minimize handoff delay without any kind of protocol standardization. We begin by presenting in Section 2, some essential technology required at suggested information based routing. In Section 3, we explain the mechanism of SNMP information based hierarchical routing. Section 4 shows minimizing scheme of handoff delay by applying new routing suggested in Section 3. In Section 5, we perform modeling, by example, about needed elements for proposed routing mechanism and legacy handoff technology to setup the routing convergence time. In Section 6, we evaluate suggested new routing mechanism accomplishing simulation for the model of Section 5 by using NS-2. Lastly in Section 7, we show the anticipated result and conclude by placing our research in the context of further study.

1888

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

2. Related works In this section, we explain underlying techniques of proposed routing mechanism, and describe how we take advantage of them. 2.1. Hierarchical MIPv6 This hierarchical MIPv6 scheme introduces a new function, the MAP, and minor extensions to the MH and the HA operations. The CN operation will not be affected. The introduction of the MAP concept minimizes the latency due to handoffs between access routers (AR). Furthermore, the addition of bicasting to a MAP allows for fast handoffs, which will minimize the packet losses due to handoffs and consequently improve the throughput of best effort services and performance of real time data services over a radio interface. Just like MIPv6, this solution is independent of the underlying access technology, allowing fast handoffs within or between different types of access networks. The introduction of the MAP concept will further diminish signaling generated by MIPv6 over a radio interface. This is due to the fact that a MN only needs to perform one local BU to MAP when changing its layer 3 access point within the MAP domain. The MAP will receive all packets on behalf of the MN it is serving and will encapsulate and forward them directly to the MN’s current address. If the MN changes its current address within a local MAP domain, it only needs to register the new address with the MAP since the global CoA does not change. This makes the MN’s mobility transparent to the CNs it is communicating with. The MAP can also be used to execute a fast handoff between ARs. When the MN uses a regional-CoA (RCoA), a MAP acts essentially as a local HA for the MN. A MAP domain’s boundaries are defined by the ARs advertising the MAP information to the attached mobile nodes. The control of a MAP’s mode of operation as an alternateCoA or a local HA is left to the network administrator’s discretion. It should be noted that the MAP concept is simply an extension to the MIPv6 protocol. Hence an HMIPv6aware MN with an implementation of MIPv6 MAY choose to use the MAP or simply use the standard MIPv6 implementation as it sees fit. Furthermore, a MN can at any time stop using the MAP. This provides great flexibility, both from a MN or a network operations point of view [4]. Fig. 1 shows composition of one WAN and two region network. There are HA and MH in Region 1, and there are one and more AR and MAP in Region 2. While CN located in WAN sends packets to MH, a MH migrates form Region 1 to Region 2 [5]. After a MH moves into

Fig. 1. A MH moves into new coverage under another map in Hierarchical MIPv6.

another link, the procedure for receiving packets form CN is as follows; 1. A MAP as router in new coverage multicasts router advertisement (RA) messages to all nodes in present link. If a MH moved into another link does not receive RA message, it multicasts route solicitation (RS) message in order to gain CoA. 2. A MH informs MAP of new CoA by sending BU message, and the MAP sends BU Ack to MH. 3. A MH informs MH of new CoA by sending BU message, and the MH sends BU Ack to MH. 4. A MH informs CN of new CoA by sending BU message, and the CN sends BU Ack to MH. 5. In case a MH moves form AR 1 to AR 2 in Region 2, a MH transmits location information only to MAP, and MAP sends BU Ack to MH. 2.2. Active network Active networks are new approach to replaces the passive packets of today’s networks with active capsule which are miniature programs that are executed at each node, such as intermediate router or switch they traverse. Accordingly, as contrasted with legacy nodes in current network that process packet header with the purpose of simple transmission between nodes, in active networks network users can make the most of node’s characteristics. So, we can make up use-driven/user-oriented network. Therefore active networks support faster service innovation by making it easier to deploy new network services and protocols without prior standardization of new service and protocols [4].

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

1889

Fig. 2. Active network architecture.

In active networks, programmable packets are called smart packet or active packet depending on approaching methodology in order to discriminate from legacy packets. Since smart packets can send both program and data in case of integrate/capsule method, they have interoperability with legacy IP networks [6]. Therefore When a packet arrives at arbitrary network node, its contents are evaluated so that the node may identify whether the packet is smart packet or not. In case the packet is identified as a smart packet. In this case, as shown Fig. 2, the capsule’s contents are dispatched to a transient execution environment of node’s operation system where they are evaluated. The scheduler transmits the evaluation results to anther active nodes or stores them in component storages which are resource repository in active node [7]. 2.2.1. Active routing As contrasted with legacy IP network in which IP packets are transmitted from source host to target node via routing path according to the routing table embedded in router, in active networks, we can take a selective routing in accordance with routing information within smart packet payload delivered by network user. Namely, active networks permit sender to inject methods which are able to invoke resident routines in an active router to modify information of routing table in an active switch node selectively. Further more, as active network allows sender to override subsistent default routing method in an active node as user-defined routing methods, so sender can reroute the path to destination according to the user-defined routing information or protocols [8]. So owing to this mechanism, we can implement new routing strategies like information based routing. 2.3. Information based routing Information based routing is with beacon routing [9] for a basis in order to transmit smart packet to destination with a high degree of efficiency. Beacons are peculiar active nodes to broadcast routing information for specific smart packets

and beacons also are operated as traditional router. On the average, active nodes are connected to one beacon or more and active packets should be transmitted to target host based on the methods within the smart packets. With the intention of deciding routing path, beacon broadcasts specific information and then sets up new link to a beacon holding target host address. Beacon routing is classified into geographic position routing, topology routing and information based routing in conformity to information of broadcasting. Since information based routing strategy takes routing procedure not according to IP address but under the information being broadcasted by another beacon and although host moves between cells, we can set up optimal path to destination. So it is very efficient to establish strategies especially for mobile environment. In information based routing, active node can select adjacent beacon based on keyword to be transmitted to the beacon. Accordingly, this means beacon connected to active node is needless to be fixed and active node can change a beacon to be connected depending on a smart packet it receives. Since routing path through which, that active packets are going is restricted within specific limit routing paths, it is so efficient that we can deliver our packets more selectively. Above all, in heterogeneous environment which active nodes coexist with legacy nodes, because active packets are transmitted just between active nodes including beacon, we can alleviate ANEP header processing time to minimum level.

3. SNMP information based routing mechanism This section presents SNMP information based hierarchical routing mechanism consisting of Keyup and Keyrout procedure. Suggested mechanism abide by the fundamental assumption of beacon-based routing. Namely, beacon is a specific active node, and each active node should be connected to one beacon or more. Also in order to make a decision routing path, beacon broadcasts specific information to all of beacons in the network. The routing

1890

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

Fig. 3. Keyup procedure.

information of routing table existed at adjacent nodes can be transmitted by the smart packets. As shown in Fig. 3, for example, by SNMP routing mechanism, service providers (source in Fig. 3) can distribute application program patches to their customers (destination 1 and 2 in Fig. 3) needing the patch fragments. For doing this, even though service providers do not know the exact IP addresses of their customer’s nodes, as service providers just send any keyword (e.g. MS_Defense) to beacon (Beacon 1 in Fig. 3), the application patch can be delivered to all of customer being in need of receiving the patch program. 3.1. Keyup procedure (keyword update procedure) In information based routing, keyword management is very significant issue since so as to send messages correctly and accurately to destination, we should take much trouble to ensure that keyword being broadcasted to all beacons should be taken charge of uniformly between source and destination node. Therefore, in this section, we suggest Keyup procedure which makes keyword information be controlled in central SNMP network management system (NMS) manager. Keyup procedure is similar to the keyword update procedure which is applied to SNMP, legacy NMS. The access management information base (MIB), hence, is accomplished by delivering SNMP protocol data unit (PDU) to manager [8], and in order to administer keyword

uniformly in centralized mode, manager has mirrorMIB storing MIB information which are changed irregularly at agent of beacons.(Fig. 4). Next steps of Fig. 3 show the Keyup procedure in stages. 1. After active nodes (Source, Destination 1, 2) create required keyword, they transmit this keyword to beacon as a smart packet encapsulating this keyword. 2. The agent of beacon makes a request for keyword authentication to manager by sending Event(keyword) message, that is, Event(1.3.1.2.1.4.20.1.6) as shown in Fig. 5. 3. After manager authenticates the keyword coming from the Beacon’s agent, it returns Set(auth_keyword) message, indeed, Set(1.3.1.2.1.4.20.1.6) to the agent of Beacon like Fig. 4 to update ipAddrTable to Beacon’s MIB. Also manager transmits Set(1.3.1.2.1.4.21) message to the agent to update ipRouteTable at beacon’s MIB to add authenticated keyword. For keeping up consistency between manager’s MIB and agents’s MIB, manager updates its mirrorMIB as identical with the beacon’s MIB information. (Fig. 5) Every time keywords of an active node may be changed, in order to achieve centralized administration for the keyword, beacon informed from the active node makes its agent deliver changed keyword to the manager by Keyup procedure.

Fig. 4. Authentication procedure between Beacon’s agent and manager.

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

1891

Fig. 5. Adding auth_keyword to ipAddrTable of MIB.

3.2. Keyrout procedure (routing procedure by keyword) Keyrout procedure in Fig. 6 is routing path establishment procedure and in this procedure, the authenticated keyword derived from previous Keyup procedure is used. This procedure, for example, has next following steps. (1) Link setup for the path, Active Node 1–Beacon 1– Beacon 4–Active Node 3 1.1 Beacons (e.g. Beancon 1, 4 and 5) set up links to each active node, e.g. source node (Active Node 1) and destination node (Active Node 3 and 4), which already created keyword and made an request for authentication for the keyword. 1.2 Beacon 1 which has been received keyword from source, Active Node 1, has got authentication over the keyword from the manager through the Keyup procedure. After doing that, Beacon 1 broadcasts the authenticated keyword to all of adjacent beacons in the network.

1.3 Beacon 4 obtaining the keyword from Beacon 1 recognizes link informations between Active Node 3 and itself. At the following step, link between Beacon 1 and Beacon 4 is established. Accordingly, full channel link is set up for the path, Active Node 1–Beacon 1–Beacon 4– Active Node 3. Through this routing path, active packets are transmitted from source to destination 1. (2) Link setup for the path, Active Node 1–Beacon 1– Beacon 2–Beacon 5–Active Node 4. Beacon 2 receiving broadcasted keyword through above (1.1)–(1.2) steps shall exchange routing table information with adjacent Beacon 5. Consequently, Beacon 2 sets up link connection to Beacon 5 after recognizing that the keyword requested by Active Node 4 is supplemented to ipRouteTable in MIB of agent of Beacon 5 through the Keyup procedure. Therefore, full channel link is set up for the path, Active Node 1–Beacon 1–Beacon 2–Beacon 5– Active Node 4 and through this routing path, active packets are transmitted from Source to Destination 2.

Fig. 6. Keyrout procedure.

1892

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

3.3. SNMP information based hierarchical routing mechanism

4. Scheme for enhancement of handoff delay diminution by using SNMP information based routing

When a MH makes soft handoff, it sends BU message. Then the BU message changes routing table of intermediate hierarchical nodes between CN and MH. In this paper, when handoffs occur continuously, we solve overhead of signaling globally, and reduce RTT and traffic of network, by configuring beacon as specific active network on intermediate router using SNMP information based hierarchical routing. We formulate the process generated by BU messages and beacon above. Let L be total number of hierarchical hops, and let M be total number of handoffs. After first handoff occurs, expression to reach BU message from MH to CN is Tint (traversing time over internet) as total number of L hops. But, message is sent only to B1, and then B1 performs optimal hierarchical routing in SNMP information based mechanism. Therefore, our proposed mechanism reduces the number of hierarchical (LK1) hops. If user configures beacon as active node to perform information based routing at intermediate nodes between MH and CN, intermediate beacons send packets based on keyword to MH through the best efficient path to HA, even though a MH moves to another area from HA area. This mechanism has an advantage that a MH receives packets from CN whenever and wherever, if a MH has known keyword provided in SNMP. Also, after a MAP manages routers and MHs in own area, and supports fast handoff in hierarchical MIPv6. But active node reduces processing time of packets, because active nodes receives keyword form SNMP and forwards packets based on the keyword globally. The active node modifies and stores binding cache with BU message. BU message does not need to be sent to HA and CN, but if a MH gets out of regional area under MAP’s management, a MH sends MU message to MAP, HA, and CN in sequence.

In this section, by trying to extend SNMP information based routing into mobile environment, we advance a suggestion about enhancement mechanism of handoff delay reduction as follows. We assume that wire node is active node or beacon and base station (BS) of new area has a connection with just one beacon. Fig. 7 shows example of network environment to explain handoff delay decreasing. 1. BS transmits advertisement messages to MH periodically. 2. In case MH leaving out from home network makes an enter into another foreign network, MH listens to default router or BS of another network in order to receive advertisement message. If MH dose not receive advertisement message within a fixed period, MH requires BS to transmit advertisement by sending solicitation to default router. 3. Depending on network prefix of advertisement message received from default router, MH makes a decision to move or not. If the network prefix is different, the mobile node assumes that it has moved. 4. If MH know that it has moved into new network, MH transmits keyword to default router as loading it with a smart packet. 5. When default router receives the keyword from MH, FA conveysthe keyword toBeacon2directlyconnectedfromit. 6. Beacon 2 delivers the keyword to adjacent Beacon 1 having the same keyword through the Keyrout procedure. 7. After Beacon 1 has got the keyword, it removes link to Beacon 3 from routing table and set up new link path to Beacon 2 which has delivered the smart packet in the previous step. Accordingly, the previous routing path was CN–Beacon 1–Beacon 3–HA–MH, but now packets are transmitted

Fig. 7. Mechanism for handoff delay reduction using SNMP information based hierarchical routing.

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

along the newly established routing path, correspondent node (CN)–Beacon 1–Beacon 2–default router–MH.

5. Modeling of enhancement mechanism for soft handoff delay using SNMP information based hierarchical routing We found mobile routing convergence time to be a principal factor for determining performance of QoS sensitive system in the view of the results of performance analysis at multimedia streaming system in mobile IP environment that was achieved so far [3]. In this section, we define routing convergence time, which means the time required for all the routers throughout the network get the same routing information. That is the elapsed time from the point when MH sends solicitation message to default router until the first packet arrives to a newly assigned address in MH after MH moved into visited network. The model is defined as follows: † Mobile node is a random walk mobility model. † Location update is on the basis of movement-based schema. † Fig. 8 shows network topology in this experiment.

1893

† Let CA be the processing cost at Router. † Let CB be the processing cost at Beacon. † Transmission cost between node A and node B is CAB. † Transmission cost is proportioned (constant proportional dT) to distance between source and destination. † Assume that distance between the wire and the node is d that is always constant. † d 0 means the distance between BS and MH. † Mobility rate of mobile node is lm. † Arrival rate of packet is la. † Processing cost at Router using new CoA is defined as CT † Traversing cost over internet, which means number of hops required for a path between source and destination is defined as Tint.

5.1. Binding update cost In mobile IPv6, MH sends BU message to HA and CN in order to solve a triangle routing problem. At the beginning, BU message is sent to HA and HA delivers packets through tunneling. At the end, MH sends BU message to CN. Therefore, we consider packets tunneling and solving triangle problem.

Fig. 8. Binding update using beacon as active node.

1894

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

Fig. 9. Process model of binding update to HA in hierarchical MIPv6.

5.1.1. Hierarchical MIPv6 In hierarchical MIPv6, there are two dependent and sequential handoffs to be considered. (1) The first handoff between HA and R1. When MH moves to R1 from HA, MH send BU message to HA. Fig. 9(a) shows BU cost between MH and HA. 3CA C 2Cmr C 2Crr This cost is identical to registration cost which MH send BU message to MAR. 3CA C 2Cmr C 2Crr

whose new name is written in parenthesis. This means that cost of the first handoff in hierarchical MIPv6 is like to one of the second handoff. Therefore, when the handoff occurs between R1 and R2, the total cost of BU is as follows. 8CA C 2CB C 5Cmr C 9Crb C Tint (3) The total cost of BU for soft handoff The total cost of BU handoff is the sum separate handoff costs (1) and (2) as follows. Binding Update cost=movement

Fig. 10(a) shows the cost, when MH send BU message to CN.

Z 2ð8CA C 2CB C 5Cmr C 9Crb C Tint

2CA C 2CB C Cmr C Crb C Cbb C Cbr C Cri C Cic C Tint Z 2CA C 2CB C Cmr C 5Crb C Tint

ðiff Crb Z Cbb

Z Cbr Z Cri Z Cic Þ Therefore, when the first handoff occurs in Hierarchical MIPv6, Total cost of BU is as follows. 8CA C 2CB C 5Cmr C 9Crb C Tint

† Let d be the average distance between HA and FA. Since transmission cost is in proportion to the d, Crb is defined as follows: Crb Z dT d

ðiff Crb Z Crr Þ

(2) The second handoff between R1 and R2 In case of the second handoff, Figs. 9(a) and 10(b) show that the Routers or Nodes are changed into ones

† In general, transmission cost to wireless links is higher than that of wire links. Assuming that transmission cost to wireless links is k times as high as that of wire links,

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

1895

Fig. 10. Process model of binding update to CN in hierarchical MIPv6.

Cmr can be defined as follows: Cmr Z kdT d 0 Therefore, average registration cost/movement may be defined as in the next expression. 2ð8CA C 2CB C ð5d C 9kd 0 ÞdT C Tint Þ Since MH’s mobility rate is lm, address registration cost is determined as in the next expression. Regmip Z lm ½2ð8CA C 2CB C ð5d C 9kd 0 ÞdT C Tint Þ 5.1.2. Handoff process model for SNMP information based hierarchical routing (1) The first handoff between HA and R1 In Fig. 11(a), the proposed procedure shows that MH forwards BU message to CN and not to HA. Intermediate beacon receives BU message, and modify its binding cache, and then sends BU Ack to MH. Therefore, the cost of BU is as follows: 2CA C CB C 2Cmr C 2Crb (2) The second handoff between R1 and R2 The second handoff refreshes binding cache of beacon located in one level up than in the first handoff. Fig. 12(a)

shows the cost of BU. 2CA C 3CB C 2Cmr C 2Crb C 2Cbb Z 2CA C 3CB C 2Cmr C 4Crb

ðiff Crb Z Cbb Þ

(3) The total cost of BU for soft handoff Consequently, average registration cost/movement is as follows: 2ð2CA C 2CB C 2Cmr C 2Crb Þ Since MH’s mobility rate is lm, address registration cost is determined as in the next expression. Regmip Z lm ½2ð2CA C 2CB C ð3d C 2kd 0 ÞdT Þ

5.2. Handoff delay and packet loss Long period of handoff delay causes mobile nodes to have high packet loss and makes network utilization to fall down. Let the packet’s arrival rate be la and assuming that the arrival of a packet is regardless of MH’s mobility velocity, then the average number of packet loss in a legacy

1896

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

Fig. 11. First handoff process model using SNMP information based hierarchical routing.

mobile IP network is defined as follows: Average number of packet drop=unit time Z la lm ½2ð8CA C 2CB C ð5d C 9kd 0 ÞdT C Tint Þ On the other hand, average number of packet drop at SNMP information based hierarchical routing procedure is shown as in the next expression. Average number of packet drop=unit time Z la lm ½2ð2CA C 2CB C ð3d C 2kd 0 ÞdT Þ

nodes are moved into new network. In case of legacy mobile IP network, Costtrans is the sum of packet transmission costs for overhead of signaling and tunneling and the optimized routing path globally. As showed Figs. 9 and 10(b), packet transmission cost in the hierarchical MIPv6 network is as follows: Costtrans Z MðCT C dT dÞ C Nð4CT C 5dT d C kdT d 0 C Tint Þ iff M O 1; N O 1 ðM is the number of handoffs; N is the number of packetsÞ

Because in the proposed mechanism packet travels are less intermediate node than in the legacy mobile IPv6 mechanism, the distance of message and data transmission is shorter, and packet is forwarded to the destination more efficiently, because the data dose not travel for Tint on topology with hierarchical structure. Therefore, above expression means that proposed mechanism reduce traffic and packet loss by solving triangle routing problem between leaf nodes and by reducing transmission cost of BU to CN.

However, in case of SNMP information based hierarchical routing procedure, Costtrans is the total amount of Beacon processing cost for the first packet and data transmission cost for the optimized routing path. As showed Figs. 11 and 12(b), packet transmission cost in SNMP information based routing is as follows:

5.3. Packet transmission to a new CoA

Considering transmission cost for Hierarchical MIPv6 network, we can find the fact that Costtrans, packet transmission cost for triangle routing, is proportioned to M and N.

In this section, we try to formulate the cost(Costtrans) for transmission packets from CA to a new CoA after mobile

Costtrans Z NðCT C 5dT d C kdT d 0 C Tint Þ iff N O 1 ðN is the number of packetsÞ

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

1897

Fig. 12. Second handoff process model using SNMP information based hierarchical routing.

On the other hand, in case of SNMP information based hierarchical routing procedure, since there is just a need for Beacon processing for the first packet and when completed, packets are transmitted through optimized path for N without additional cost for tunneling, we can cut down the cost considerably in comparison with legacy mobile IP network mechanism.

6. NS-2 simulation In this section, by using a network simulator, we generate the scenario and simulate an existing simple handoff and proposed handoff procedure modeled in Section 5. System environment for performance evaluation are as shown in Table 1.

freely within a bounded area, and global mobility moving from one side of the topology to another. Also it has embedded extensions to allow simultaneous use of wireless features, mobility, and multicast. It contains library to create, configure, and manipulate large topologies in a very easy manner, and translator form GT-ITM to NS-2 which output TOPOMAN procedure calls. (2) Information based active network For active networking simulation in ns, TASC developed the AN package in the PANAMA project. But the PANAMA’s AN package for ns is quite specific to reliable multicast or their own protocol, AER and it requires modifying the classifier. So it is difficult to apply to a generic implementation of active networking. Therefore, to be able to simulate the proposed handoff optimization that is based on active networking, ns had to be extended for the generic context of active networking and evaluation of

6.1. NS-2 extension (1) Mobiwan Mobiwan has been developed by MOTOROLA labs in collaboration with INRIA PLANETE Team. Mobiwan is a simulation tool based on NS-2 meant to simulate mobile IPv6 under large wide-area network, local-area mobility, and global-area mobility. And it supports simple IPv6 extensions needed by mobile IPv6, local mobility moving

Table 1 System environment Hardware

Pentium-IV 1.6 GHz CPU

Operating System Software

FreeBSD 4.7 NS-2 2.1b6 (Berkeley wireless extension [8]), Mobiwan, Gnuplot CCC, Otcl, Perl, awk

Programming language

1898

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

Fig. 13. Definition for Asnmp agent.

the proposed handoff optimization procedure. In this purpose, the following modules were added:

(1) Asnmp agent To exchange the packets between nodes, agents have to attach to the nodes. For enabling information based active network packets to be transmitted and executed, we define the Asnmp Agents based upon the fact that Smart Packet’s contents are dispatched to a transient execution environment in node OS. Asnmp Agent is implemented as a subclass of Agent/Message as shown in Fig. 13.

internet protocol FTP and exponential traffic in application. The different between legacy mechanism and proposed mechanism weather Asnmp agent is used or not used to send smart packet. Our simulation is based on the assumption that MH receives the certified keyword before moving. Therefore, after MH moves to another area, MH forwards a smart packet containing a keyword, because MH already took the certified keyword before moving. Fig. 14 shows topology creation and simulation scenario based on movement pattern specification. Intermediate W(1) and W(2) operate as MAP1 and MAP2, respectively, in legacy method, they performs roles as Beacon1 (Asnmp1) and Beacon2 (Asnmp2) individually.

6.2. Simulation scenario

6.3. Evaluation of the optimized handoff protocol

In order to investigate the impact of the proposed handoff optimization on throughput and network traffic, the comparison target is throughput of TCP packet using

We transformed a subset of the trace data of interest into Gnuplot input format using Perl. Among the trace date, we selected the column in order to focus on:

† New Ee agent † New handoff procedure

Fig. 14. Generation of topology and movement pattern.

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

† Throughput comparison † Traffic analysis † Packet transfer delay

6.3.1. Throughput comparison Fig. 15 shows TCP packet throughput between CN and MH during simulation. And the result of Fig. 15 is simulated on 1000 kB and 10 MB data transmission. When a MH changes HA to default router of new MAP area, BU message has to be registered with both HA and CN, and to be sent to CN through the number of N hops between MH and CN. But BU message does not have to reach HA and CN by using SNMP information routing mechanism proposed in this paper, even though a MH switches MAPs. Therefore, we find out TCP throughput improvement in proposed mechanism than hierarchical MIPv6 from Fig. 15. Because a MH sends BU message only to MAP even if it roams in MAP’s area, the number of transmission hops in hierarchical MIPv6 is same to that in SNMP information based hierarchical routing. But processing time of active node is more efficient than processing time of MAP for IP based transmission, because packets are transmitted on the basis of keyword in case of information based hierarchical routing. Recently, most people want to receive many data to handset, because they desire to do business works and to see moving pictures, etc. The result of Fig. 15 shows that our proposed mechanism is of benefit to process many packets on real time. Also, although data transmission changes 1000 kB to 10 MB, 10 MB packet throughput of proposed mechanism is higher than 1000 kB packet throughput of Hierarchical HIPv6. In Fig. 15, when a MH moves from HA to new MAP’s area at before and after 20 s, throughput of Hierarchical

1899

MIPv6 is increased by BU messages and handoff delay, but throughput of SNMP information based hierarchical routing is decreased by reducing BU message. In case a MH changes BS with another BS in same MAP’s area, throughput of proposed mechanism is a litter higher than throughput of legacy mechanism. But when a MH moves to coverage of another MAP, the gap between two throughputs gets bigger by handoff delay at before and after 80 s. The legacy mobile IPv6 has M(CTCdTd) has a more higher registration cost than the proposed mechanism in Section 5. Packet processing cost and transmission cost increase whenever M (handoff) occurs. The result graph shows this cost difference between legacy mobile IPv6 and proposed mechanism through difference of throughput before and after 20 and 80 s when MH performs handoffs. This result graph of throughput is affected by difference of registration cost, that is, 6CAC(2dC7kd 0 )dTCTint. In case of SNMP information based hierarchical routing procedure, because there is just a need for Beacon processing of the first packet and once completed, packets are transmitted through optimized path for N without additional cost of tunneling. This analysis also affects traffic result of next sub section. 6.3.2. Traffic analysis In this sub-section, we analyze the network traffic generated by applications. A rapid growth of the Internet and proliferation of new multimedia applications lead to an exponential growth in the volume of application traffic over IP networks. Traffic occurs whenever intermediate nodes process packets which become an exponential growth. Fig. 16 shows that traffic of proposed mechanism is decreased than traffic of legacy mechanism on 1000 kB and 10 MB data transmission. The reason is that network overhead is reduced by decreasing BU messages both from MH to HA

Fig. 15. Comparison result of throughput.

1900

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

Fig. 16. Comparison result of traffic.

and from MH to CN. Our proposed mechanism enables beacon to process packet faster. Therefore, it optimizes transmission path and reduces delay time using active node and information based routing after a MH receives new CoA in switching area. Reduction of traffic overload over network is very important, because more data packets are transmitted to handset, more unnecessary signaling and traffic get larger. In Fig. 16, when data transmission changes 1000 kB to 10 MB, network traffic is raised in both proposed mechanism and Hierarchical MIPv6. But an advantage of proposed mechanism is that traffic of proposed

mechanism for 10 MB data transmission is lower than traffic of Hierarchical MIPv6 for 1000 kB data transmission. So when many data packets are transmitted over network, we find out that our proposed mechanism is more efficient than Hierarchical MIPv6. When a MH moves from coverage of HA to coverage of new MAP at before and after 20 s, the gap between one traffic of hierarchical MIPv6 and the other traffic of SNMP information based hierarchical routing gets bigger as seen in Fig. 16. In case a MH changes BS with another BS in same MAP’s area, traffic of proposed mechanism is a litter higher than that of legacy mechanism. But traffic of hierarchical

Fig. 17. Comparison result of packet transfer delay.

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

MIPv6 is increased, and traffic of proposed mechanism is decreased by handoff delay at before and after 80 s. 6.3.3. Data transfer delay Data transfer delay makes it difficult to support real time service and multimedia streaming service by interfering with transmitting packets fast from CN to MH. In particular, the core requirements of real time service have something to do with packet delay, especially, have a close relation to hand off delay. Moreover, data transfer delay is more important at wireless mobile environment. In Fig. 17, packet delay of Hierarchical MIPv6 is higher than that of proposed mechanism. When first handoff occurs, the packet delay is higher in Hierarchical MIPv6 at before and after 20 s. But proposed mechanism shows a little packet delay at before and after 20 s by setting up transmission path with keyword in active node first. Also, when a MH makes another handoff which a MH moves into coverage of another MAP, the packet delay is higher in Hierarchical MIPv6 at before and after 80 s, but packet delay in proposed mechanism is lower. As seen in the graph roughly, we find out that packet transfer delay for proposed mechanism is lower than Hierarchical mechanism, and delay value is regular roughly until simulation is finished. This means that proposed mechanism decreases data transfer delay by reducing unnecessary signaling and network overload. Therefore, a handset receives real time services and multimedia services over wireless mobile network in large quantity. In Section 5, we see that the registration cost, 6CAC (2dC7kd 0 )dTCTint, occurs, when MH performs handoffs. That means that the processing cost of router and transmission cost in wired and wireless area is increased. Moreover, in case that MN communicates with a CN at a far distance, Tint cost effects throughput. Therefore, when MH handoffs twice, registration cost is 2(6CAC(2dC7kd 0 )dTC Tint) and traffic result of legacy mobile IPv6 is higher than one of the mechanism proposed in this paper. That is because, in the proposed mechanism packet travels through less intermediate nodes than in the legacy mobile IPv6 mechanism, the distance of message and data transmission is shorter, and packet is forwarded to the destination more efficiently, since the data dose not travel for Tint on topology with a hierarchical structure. 6.3.4. Entire comparison We calculate throughput and traffic from data gained during simulation. Table 2 shows comparison evaluation of entire data values calculated for simulation. As shown in Table 2, for throughput, proposed mechanism processes 204.681 Byte/ms more much on 10 MB data transmission and 202.164 Byte/ms more much on 1000 kB data transmission than legacy mechanism, respectively. Therefore, proposed mechanism enhances 13.23% on 10 MB data transmission and 12.01% on 1000 kB data transmission as a whole. For traffic, proposed

1901

Table 2 Comparison for HMIPv6 and SNMP information based hierarchical routing

10 MB Throughput 1000 kB Throughput 10 MB Traffic 1000 kB Traffic Total Delay

Legacy mechanism

Proposed mechanism

Comparison (enhance) (%)

1298.432

1503.113

13.23

1398.938

1601.102

12.10

1.34716 1.26716 9.037608

1.194318 1.128152 7.695045

11.34 10.97 14.85

mechanism reduces 0.152842 Byte/ms on 10 MB data transmission and 0.139008 Byte/ms on 1000 kB data transmission, respectively. Hence, 11.34% on 10 MB data transmission and 10.97% on 1000 kB data transmission is improved all over. For data transfer delay, we find out enhancement as 14.85% in proposed mechanism, because 1.342563 Byte/ms is decreased. This enhancement is sure to result from reducing data traffic and packet loss, because signaling for informing own location is decreased, when a MH makes softer handoff. Also optimization of signaling by using active node causes traffic diminution and throughput augmentation, because the signaling optimization reduces the procedures for sending many data to MH.

7. Conclusion The focus of our research is to make QoS guaranteed in mobile IP. QoS in mobile IP is important to provide multimedia and real-time applications services in mobile environment, and it is closely related to the handoff delay. Therefore, handoff delay problem is actively studied to guarantee QoS in mobile IP research area. In this paper, to resolve such a problem, we suggest SNMP information-based hierarchical routing that adds keyword management method to Information-based hierarchical routing in active network, and then we suggest QoS controlled handoff by SNMP information-based hierarchical routing. To prove the QoS improvement in our approach, we model the suggested handoff and simple handoff after setting up routing convergence time, and then simulations were carried and using NS-2. The simulation results of handoff optimization that was proposed in this paper show that it improved throughput and reduced application traffic avoiding handoff delay caused by triangle routing problem. Therefore, the result of experiment validate that propose mechanism performs fast handoff and improves QoS by reducing handoff delay. As a further research, we are trying to improve QoS of packet that is transmitted through both hierarchical wired nodes and hierarchical wireless nodes using proposed mechanism in this paper, when ad hoc network forms with many mobile nodes.

1902

S.-H. Ryu / Computer Communications 28 (2005) 1887–1902

References [1] K.Y. Ko, J.K. Kim, Mobile IP improvement for micro mobility support, Kiss, Proceeding of 2001 Fall. [2] S. Schemid, J. Finney, A. Scott, D. Shepherd, Active component driven network handoff for mobile multimedia system, Proceedings of the Seventh International Workshop on Interactive Distributed Multimedia Systems and Telecommunications (IDMS), October 2000, 2000. [3] D. Chalmers, M. Sloman, A survey of quality of service in mobile computing environment, IEEE Communication Surveys Second Quarter (1999). [4] H. Soliman, et al., Hierarchical MIPv6 mobility management, draftietf-moblieip-hmipv6-02.txt, February 2001. [5] E. Kupiainen, Hierarchical Mobile IPv6 Mobility Management, February 2003. [6] D.S. Alexander, B. Braden, C.A. Gunter, A.W. Jackson, A.D. Keromytis, G.J. Minden, D. Wetherall, Active Network Encapsulation Protocol (ANEP), RFC, ANEP Documentation, April 1997. [7] D.L. Tennehouse, D.J. Wetherall, Toward an active network architecture, ACM Computer Communication Review 26 (2) (1996) 5–18.

[8] A.B. Kulkarni, G.J. Minden, Active networking services for wired/wireless networks, INFOCOM’ 99 1999;. [9] http://www.ittc.ku.edu/~ananth/845.html, Beacon Routing in Active Network.

Further Reading [1] D.L. Tennenhouse, J.M. Smith, W.D. Sincoskie, D.J. Wetherall, G.J. Minden, A survey of active network research, IEEE Communications Magazine 35 (1) (1997) 80–86. [2] A. Mark, P.E. Miller, Inside Secrets SNMP Managing Internetworks, SamGakHyung Press, 1998. [3] W. Yu, C. Weidong, S.M.H. Joseph, Performance Analysis of Adaptive Location Management for Mobile IP, Technical Report 97-CSE-13, Southern Methodist University, 1997. [4] The CMU Monarch Project’s Wireless and Mobility Extension to ns, The CMU Monarch Project, August 1999. [5] K. Anil, Gopinath, Implementing New Internet Services using an Active Network.