Additional Performance Evaluation Results of the Proposed NFHO Scheme

With different transmission delays, α, we first investigate the relationship between the expected number of buffered packets in the target BS and the packet arrival rate. We assume random variables, µ, ν and α, have mixed Erlang-2 distributions [18][19][21]. The coefficients for µ, ν, and α were set to µ₁ = µ₂ = 0.5, ν₁ = ν₂ = 0.5, and α1 = α2 = 0.5, respectively [18]. Other parameter settings are shown in each figure. In addition, the mean time of HO message processing delay is computed as α = 2(α₁ σ₁+α₂ σ₂). By Eq. (10), Fig. 3.13 shows the relationship between the expected number of buffered packets in the target BS and the packet arrival rate under three different α . It was observed that the expected number of buffered packets (Qmax) is positively correlated to the packet arrival rate (λ) of the MS. An increase in λ leads to a larger Qmax, and a larger α also results in a larger Qmax. Note that even with other parameter settings, we still can obtain similar results.

Since the size of the HO packet buffer pool is limited, the packet loss probability during HO is affected by the packet arrival rates of ongoing HO MSs and the available packet buffers in the pool. We consider three different sizes of the HO packet buffer pool, and the BS can only support a limited number of concurrent HO requests. By Eqs. (18), (21), and (26), we evaluate the relationship among C (the allowance of concurrent HO MSs), M (the size of the HO packet buffer pool), and Ploss (packet loss probability). Fig. 3.14(a) shows that the packet loss probability (Ploss) of an HO MS increases with the increase of the allowance of concurrent HO MSs (C). The BS that provides a larger size of HO packet buffer pool (M) has a lower packet loss probability during HO. Note that Ploss approximates zero when C and M are set to 5 and 300, respectively. This is because M is large enough to accommodate all arrival packets from C HO

and M. The parameter settings, listed on top of Fig. 3.14(a), will also be used in Fig. 3.15, Fig.

3.16, and Fig. 3.17, unless there are individual settings listed in each figure.

Size of the HO packet buffer pool, M 0

1 2 3 4 5 6 7 8 9 10 11 12

50 100 150 200 250 300 350 400

Ploss=0.3 Ploss=0.15 Ploss=0.05

Fig. 3.15: The relationship between M (the size of the HO packet buffer pool) and C (the maximum allowance of concurrent HO MSs) under a given Ploss (packet loss probability)

constraint.

Packet loss probability is an important QoS parameter for HO. According to the size of the HO packet buffer pool allocated in a BS, the BS can control the allowance of concurrent HO MSs to meet the packet loss probability requirement. With a given packet loss probability, by Eqs. (18), (21) and (26), we evaluate the relationship between C (the allowance of concurrent HO MSs) and M (the size of the HO packet buffer pool). Fig. 3.15 shows that when allocating a larger M, the BS can accept a higher C while still meeting the packet loss probability constraint.

With a given Ploss, Fig. 3.16 also shows that the higher C that the BS allows, the lower λ (packet arrival rate) that the BS can support to each MS. Note that, without loss of generality, the ξ /ε ratio was set to 20, and M was set to 300. It can be observed that when we increase C, the supported λ of each MS decreases. The decreasing of λ is sharp until C is greater than 20. It is because the expected value of packet holding time grows until C exceeds the ξ /ε ratio. Similar

results are also shown in Fig. 3.17. Given three different ξ /ε ratios, it can be seen that when we increase C, Ploss grows sharply until C is greater than the corresponding ξ /ε ratio. Fig. 3.16 and Fig. 3.17 also show that the BS can guarantee the packet loss probability requirement for ongoing HO MSs by controlling the allowance of concurrent HO MSs (C) and the ξ /ε ratio.

Therefore, to provide proper QoS (e.g., the packet loss probability, Ploss) to ongoing HO MSs, Eqs. (18), (21) and (26), which were the basis of Fig. 3.16 and Fig. 3.17, can assist to setup an appropriate admission control policy to grant or reject incoming HO requests.

(/sec)

Allowance of concurrent HO MSs, C

Packet arrival rate, λ

0 50 100 150 200 250

5 10 15 20 25 30 35 40

Ploss < 0.05

=20

ε ξ

=300 M

Fig. 3.16: Packet arrival rate (λ) of each MS versus the allowance of concurrent HO MSs (C) under a givenpacket loss probability (Ploss).

=30

ε ξ

=10

ε ξ ε

ξ

Fig. 3.17: Ploss versus C under a given ξ /ε.

Chapter 4

A Dynamic MBS Zone Framework for Cost-Effective Inter-MBS Zone

Handover in WiMAX Networks

To synchronize MBS (multicast broadcast service) zone data transmission, the WiMAX standard defines a coordination mechanism to coordinate data transmission over the WiMAX network; however, the packet loss recovery procedures, which are parts of the coordination mechanism, enlarge the packet transmission latency and packet buffer pool requirement. In this Chapter, we propose an in-frame control (IFC) scheme to decrease the packet error rate and packet retransmission count, so as to reduce the packet transmission latency and packet buffer pool requirement. Furthermore, to support level-2 frame-offset coordination, we also propose a dynamic MBS zone (DMZ) framework that can provide data continuity between any two

adjacent MBS zones. Based on the proposed DMZ framework, a seamless dynamic inter-MBS zone handover (called DMZ HO) scheme is proposed to resolve the data discontinuity (or packet

loss) problem during inter-MBS zone HO. An analytic model has been developed to analyze the packet error rate and packet retransmission count for the proposed IFC scheme and the bandwidth overhead in terms of channel occupation time for the proposed DMZ HO.

Performance evaluation results show that, compared to the original WiMAX scheme, defined in the WiMAX standard, the proposed IFC scheme reduces the packet error rate and packet

0.001 (/ms)), respectively. Moreover, the proposed DMZ HO scheme outperforms an existing overlapping zones (OLZ) scheme. It reduces channel occupation time by 89.3% compared to the OLZ scheme when the HO arrival rate, µ, is 1 (/sec). In addition, the proposed DMZ HO scheme consumes less channel bandwidth than the OLZ scheme. The channel idle ratio of the proposed DMZ HO scheme is 89.3% (µ = 1, PDU-offset = 10) larger than that of the OLZ scheme.

Therefore, the proposed DMZ HO scheme is more cost-effective than the OLZ scheme and is thus very feasible for interactive TV (ITV) applications.

4.1 Proposed DMZ HO

The level-2 frame-offset coordination requires the data transmissions between any two adjacent MBS zones are synchronized so that it can provide HO MSs with the continuity of data reception during inter-MBS HO [23]. Therefore, during inter-MBS zone HO, providing data continuity for HO MSs is equivalent to support the level-2 frame-offset coordination. In this section, we propose a DMZ framework to support level-2 frame-offset coordination for inter-MBS zone HO.

Based on the proposed DMZ framework, a seamless DMZ HO scheme is proposed to resolve the positive PDU-O value problem between any two adjacent MBS zones so as to support HO MSs for data continuity during inter-MBS zone HO.

4.1.1 Aggregating MBS Sync Rules and MBS Payload

For transmission synchronization, the MBS sync rules distribute the transmission information of corresponding MBS payloads. Therefore, any packet error of MBS payloads or their associated sync rules during the Accumulation Period should be recovered within the following Recovery Period. The Period should be long enough to finish both sync rule and data path recovery procedures. A longer Recovery Period means a larger size of a packet buffer pool required, and

it results in longer transmission latency. Reducing the packet loss probability of both MBS payloads and their associated sync rules can lead to lower transmission latency and a smaller size of the packet buffer pool. In this Chapter, we propose an in-frame control (IFC) scheme that aggregates the MBS payload and its associated MBS sync rules. In contrast to transmitting the MBS payload and its associated MBS sync rules separately, binding both the MBS payload and its associated sync rules together can reduce the probability of starting recovery procedures. A packet aggregating an MBS payload and its associated sync rules is termed as an IFC packet. By using IFC packets instead of individual MBS payload packets and sync rules, we can shorten the packet transmission latency and reduce the size of the packet buffer pool.

Fig. 4.1: Proposed DMZ HO.

4.1.2 Dynamically scaling up the MBS zone

The HO process of the IEEE 802.16-2009 [1] can be functionally divided into two stages: HO preparation and HO execution. In the HO preparation stage, an MS selects a target BS and then

proceeds with HO initialization. In the HO execution stage, the MS performs downlink synchronization, ranging and network re-entry; this stage starts actual HO. Based on the HO process, we propose a DMZ framework to support level-2 frame-offset coordination. The proposed DMZ framework dynamically overlaps the serving and target zones to resolve data

discontinuity. After the HO is completed, the serving and target zones are reverted to non-overlapping. Without loss of generality, we assume that each MBS zone is a single-BS MBS zone. Fig. 4.1 shows an overview of the proposed DMZ HO. In the proposed DMZ framework, the inter-MBS zone HO is divided into three phases: source zone extension, intra-MBS zone HO, and intra-cell inter-MBS zone HO. The source zone extension and

intra-MBS zone HO operate in the HO preparation stage and HO execution stage respectively.

The intra-cell inter-MBS zone HO is started when the MS resolves the data discontinuity between the serving and target MBS zones. Assume that the MS is now served by the serving MBS zone. In the HO preparation stage, the serving ASN-GW performs the source zone extension by adding the target BS to the serving MBS zone. After the target BS joins the serving MBS zone, the serving MBS zone becomes a two-BS MBS zone, and the serving IFC packets are delivered to the serving BS and the target BS. Then, in the HO execution stage, an MS treats the HO within the serving MBS zone as an intra-MBS zone HO. In the HO execution stage, the MS handovers to the target BS and then joins the target MBS zone. As to the serving MBS zone, the MS performs an intra-MBS zone HO to switch the MBS services from the serving BS to the target BS. In the intra-MBS zone HO, the frame-level coordination is used for data synchronization. At the beginning of the intra-cell inter-MBS zone HO, the MS not only belongs to the serving MBS zone but also join the target MBS zone. Through the target BS, the MS concurrently receives the MBS payloads from both the serving MBS zone and the target MBS zone until redundant MBS payloads are received from either zone. Note that in the case of negative PDU-O value, the MS will immediately detect redundant MBS payloads from the target MBS zone. On the contrary, in the case of positive PDU-O value, the MS will not detect redundant MBS payloads until all the PDU-O packets are received from the source MBS zone.

After detecting redundant MBS payloads, the MS informs the target BS to terminate the serving

Fig. 4.2: An illustration of the proposed DMZ HO. (a) An MS moves from MBS zone1 to MBS zone2. (b) Perform source zone extension to add BS2 to zone1. (c) The MS handovers to

BS2 by performing intra-MBS zone1 HO, and then the MS also joins zone2. (d) The MS performs intra-cell inter-MBS zone HO in BS2, and then BS2 is removed from zone1.

An illustration of the proposed DMZ HO is given in Fig. 4.2. Fig. 4.2(a) shows that both MBS zone1 and MBS zone2 provide the same MBS service. ASN-GW1 and ASN-GW2 are in charge of zone1 and zone2, respectively. When the MS in BS1 decides to handover to BS2, ASN-GW1 adds BS2 to zone1. Fig. 4.2(b) performs the source zone extension so that BS2 joins zone1, and ASN-GW1 is the anchor ASN-GW of zone1. At this moment, BS2 provides both zone1 service and zone2 service. The zone1 IFC packets are issued from the ASN-GW1 (anchor) to both BS1 and BS2. Moreover, BS2, providing zone2 service, still keeps receiving IFC packets from zone2. Fig. 4.2(c) shows that the MS performs intra-MBS zone HO in the zone1, and it also joins the zone2. The MS, simultaneously receiving MBS bursts from both zone1 and zone2

through BS2, checks for redundant payloads by inspecting the sequence number of MBS payloads from both zones. Upon receiving redundant payloads, the MS issues the intra-cell inter-MBS zone HO by signaling BS2 to terminate the zone1 service and then receiving MBS service dedicated from zone2. Fig. 4.2(d) shows that the MS completes the intra-cell inter-MBS zone HO, and the target BS is removed from zone1.

#2

Intra-MBS zone HO(HO execution) Source zone extension(HO preparation) Intra-cell inter-MBS zone HO

MOB_HO-IND

DL Synchronization

RNG-REQ / CDMA ranging code RNG-RSP(SBC-RSP, REG-RSP)

Release Serving Zone Context

RNG-REQ / CDMA ranging code MOB_NBR-ADV

Serving MBS Zone Packets

Message relays

via ASN-GWs Direct Message Target MBS

Zone Packets

Fig. 4.3: MSC of the proposed DMZ HO.

Based on Fig. 4.2, Fig. 4.3 shows the detailed message sequence chart (MSC) of the proposed DMZ HO. BS1 and ASN-GW1 are the serving BS and serving ASN-GW, respectively.

BS2 and ASN-GW2 are the target BS and target ASN-GW, respectively. The bold lines and dashed bold lines show the MBS IFC packet flows to the serving MBS zone and target MBS zone, respectively. The dashed thin lines are management messages. The solid thin lines are sent directly from source to destination, and the dashed thin lines sent from source to destination are relayed via the two ASN-GWs. The management messages termed with all capital letters are the MAC management messages defined in the IEEE 802.16-2009 [1], and the other management messages termed with leading capital letters and followed by low case letters are the control messages defined in the WiMAX standard [23]. The proposed seamless inter-MBS zone DMZ HO scheme based on the DMZ framework is detailed as follows (see Fig. 4.3):

1) The MS gets adjacent BSs information from the received MOB_NBR-ADV message. The MCID pre-allocation information in the MOB_NBR-ADV message provides MCID mappings between the serving MBS zone and adjacent MBS zones. In the HO preparation stage, the MS starts scanning by issuing an MOB_SCN-REQ message to request a group of time intervals from the serving BS (BS1), and the serving BS grants the time intervals by replying an MOB_SCN-RSP message. Within the time intervals, the MS associates to BS2, and eventually the serving BS sends association results to the MS by sending an MOB_ASC-REP message.

2) The MS issues an MOB_MSHO-REQ message to notify that BS2 is the HO target. The serving BS sends an HO_Req message to the target BS (BS2). The HO_Req message which carries MBS service flow information will enable the target BS to start creating an MBS service flow to the serving ASN-GW1.

3) After inspecting the information in the HO_Req, the target BS initiates a dynamic service flow creation for the MBS. By exchanging the Path_Reg_Req/Rsp/Ack and

RR_Req/Rsp/Ack messages, the data path between the serving ASN-GW1 and the target BS is created, and then IFC packets from the serving MBS zone are delivered to the target BS. At the end of the HO preparation stage, both the serving and target zones are overlapped at the target BS.

4) In the HO execution stage, the MS synchronizes to the target BS DL. By obtaining the MCID pre-allocation information from the MOB_NBR-ADV message, the MS joins the target MBS zone and is able to receive both the serving and target MBS packets simultaneously before starting the ranging and network re-entry. As shown in Fig. 4.3, after DL synchronization, the MS receives packet #1 from the serving zone and packet #2 from the target zone before the RNG-RSP message is received.

5) After completing the ranging and network re-entry, the MS keeps receiving MBS bursts from both the serving zone and target zone until a redundant MBS payload is found. Note that the redundancy check can be done by checking the sequence number, such as the transport layer sequence number in the reassembled MBS bursts. Once a redundant MBS payload is found, the MS issues HO_mbs_switch to notify the target BS that the MS will leave the serving MBS zone. As shown in Fig. 4.3, the MS receives a packet with sequence number #2 from the target zone before issuing an RNG-REQ message. After that, another packet with the same sequence number #2 is also received from the serving zone. The MS assures that data discontinuity is resolved and then issues HO_mbs_switch to finalize the inter-MBS zone HO.

Note that the HO case shown in Fig. 4.3 assumes that the PDU-O value is 1 between the source and target zones.

6) Any redundant MBS payload from the serving or target zone indicates that the HO has resolved the potential loss problem of MBS payloads. After receiving an HO_mbs_switch message, the target BS might delete the service flow by exchanging

from the serving zone is issued only when there is no subsequent ongoing HOs. After this procedure, the target BS leaves the serving zone, and now the service and target zones are non-overlapped.

4.2 Analytic Model

In this section, we will analyze the performance of the proposed IFC scheme and DMZ HO scheme. The proposed IFC scheme focuses on preventing the BS from entering the recovery procedures. Therefore, the packet error rate and packet retransmission count are two key performance metrics. In section 4.2.1, we analyze the packet error rate and the expectation of the packet retransmission count. In addition, the proposed DMZ HO scheme supports data continuity by dynamically allocating a channel during HO to transport PDU-O between the serving MBS zone and the target MBS zone. The bandwidth overhead of an HO can be reflected by the usage of the dynamic allocated channel. In section 4.2.2, we analyze the channel usage in terms of the expectation of the channel occupation time.

Fig. 4.4: State transition diagram of the Markov process for a link.

4.2.1 Derivations of the Packet Error Rate and the Expectation of the Packet Retransmission Count

To analyze the packet retransmission count, we assume the arrivals of physical impairments in a link are a Poisson process [26]. A physical impairment results in a period of incorrect transmission, which was termed as an error cluster. Depending on the nature of physical impairments, the durations of error clusters also have a probability distribution [26]. In addition, an error only occurs within an error cluster, and the last bit of each error cluster is always an error bit [26].

We model a link with errors by the M/M/1/1 queuing model and assume a link is either in the good state (GD) or in the error burst (EB) state [26]. The GD state denotes a link is out of error clusters durations, and the EB state denotes a link is within the periods of error clusters. Fig.

4.4 shows the state transition of the Markov process for a link. Assume that the arrivals of error clusters are a Poisson arrival process with arrival rate λ, and θ denote the departure rate of the EB state. Thus, the inter-arrival time is exponentially distributed with the probability density function [17][18][21].

( ) t = e

t ≥ ⁰

f λ

Let δ denote a random variable to represent the duration of an error cluster. Assume that δ has the

Let δ denote a random variable to represent the duration of an error cluster. Assume that δ has the