Related Work on Queueing Analysis for Wireless Transmission Systems

In general, in a wireless transmission system, arriving packets are buffered into radio link level queue before transmission to the target mobile (in downlink) or to the base station (in uplink). This buffering process causes delay in wireless transmission. Again, the transmission rate of a wireless channel varies due to variations in channel quality and channel errors. Therefore, a queueing analysis which can capture the channel and the radio link level buffer dynamics becomes a useful mathematical tool to investigate the QoS performances of a wireless transmission system. By using a queueing model, the performance results can be obtained more efficiently when compared with simulations. In addition, a queueing analytical model can be used not only to analyze the system's behavior under different parameter settings, but also to optimize the system performances in which several standard techniques in optimization (e.g., Markov decision process) can be applied.

Different queueing models for wireless transmission systems were proposed in the literature. Queueing analyses for a polling system and a system with cyclic service queues with Bernoulli scheduler were presented. These works considered only single rate transmission at the physical layer in which the modulation and coding scheme is fixed. On the other hand, multirate transmission based on adaptive modulation and coding (AMC) has been proposed in most of the current wireless standards to archive higher system capacity. Queueing analysis for radio link level scheduling with adaptive modulation in time division multiple access (TDMA) system was proposed. Also, a queueing model was used for optimizing the radio link level system parameters.

Multimedia services will be common in next generation wireless systems. A queueing model developed specifically for video source (e.g., MPEG traffic) was presented. In addition, the authors demonstrated how the video source coding parameters can be optimally adjusted to achieve the best performance. Again, in a multiservice traffic scenario, prioritization of real-time traffic over best-effort traffic is required. An analytical model for priority queueing was presented.

Combatting transmission errors due to interference and noise is one of the challenging issues in wireless transmission systems. Automatic Repeat reQuest (ARQ) is one of these methods to recover erroneous transmissions. When a transmission fails (i.e., the receiver cannot decode the transmitted information correctly), the receiver requests the transmitter to retransmit. Different variants of ARQ can be used (e.g., probabilistic retransmission, finite and infinite retransmission). Queueing models with ARQ mechanism were proposed. For example, queueing models for go-back-N and selective repeat ARQ were presented. Since with AMC multiple packets can be transmitted in one time slot, if an error occurs, N packets up to the last transmission will be retransmitted for go-back-N ARQ and only the erroneous packets will be retransmitted in case of selective repeat ARQ. Due to the time and space-dependent wireless channel errors and the burstiness of the errors, some connections could experience inferior performance compared with the others. Therefore, a compensation mechanism was introduced to maintain fairness (by allowing more transmissions in the current time frame to compensate errors in the previous time frame) and guarantee the target QoS performance. Also, a queueing model for this compensation mechanism was proposed.

In a multi-user wireless system, packet scheduling is required to allocate the available transmission resources to the ongoing users in a fair and efficient manner. Various packet scheduling policies were proposed in the literature. The most common policy is fair scheduling in which the ongoing users receive services based on their preassigned weights. A queueing model for this scheduling policy in a wireless transmission environment was presented. On the other hand, opportunistic scheduling was developed specifically for wireless transmission systems. This scheduling policy takes advantage of multi-user diversity to improve the throughput of the entire system. In particular, the user who has the best channel quality in the current time slot will be selected to transmit. A queueing model of this scheduling policy was proposed. In addition, queueing analyses for different resource-sharing schemes (i.e., max-min fairness, proportional fairness, and balanced fairness) were presented. Also, the stability condition for each of the schemes was studied.

While most of the queueing models in the literature considered only the variation of wireless channel on system performance, a few works considered the impacts of resource allocation and admission control on queueing performance. For example, impacts of resource allocation and admission were considered in the queueing model for a TDMA-based cellular wireless system  using adaptive modulation and coding, and a code division multiple access (CDMA)-based system  with rate adaptation. These investigations showed that resource reservation for handoff connections (i.e., through guard channel) as well as the transmission rate adaptation can impact the queueing performance of the mobile users significantly.

Queueing analysis can be also used for the development of admission control mechanisms. This approach was presented in the literature. In particular, given the traffic parameters and the estimated wireless channel quality, for a given medium access control mechanism, information on queueing delay and packet loss can be obtained and used by the admission controller to decide whether a new connection can be accepted or not. The decision on acceptance or rejection of a new connection is based on whether the QoS performances of both the ongoing connections and the new connection can be maintained at the target level or not. A queueing model was presented which could be used for admission control in Bluetooth-based wireless personal area networks (PANs). A queueing model for IEEE 802.11-based wireless local area networks (WLANs) was proposed. In these works, the MAC protocol was assumed to be carrier sense multiple access/collision avoidance (CSMA/CA).

In addition to the traditional wireless systems which rely on the Single-Input Single-Output (SISO) transmission, Multiple-Input Multiple-Output (MIMO) system has been developed to provide better performance in terms of error or transmission rate. A queueing model for MIMO system was proposed in which the advantages of spatial diversity in terms of smaller queueing delay and packet loss were demonstrated.

Multihop communication will be a significant component in the next generation wireless systems. In a multihop network, the transmission range of a transmitter (e.g., in the base station) can be extended by relaying the traffic through multiple intermediate nodes. This multihop transmission is a common feature in wireless ad hoc, mesh, and sensor networks. Queueing models for this multihop wireless communication were proposed in the literature. For example, end-to-end performances of a multihop wireless network in terms of latency, reliability, and throughput were studied through a queueing model considering adaptive modulation and coding at the physical layer. A tandem queueing model for multihop transmission was presented in. For sensor networks, energy conservation is one of the challenging issues. Since the amount of energy available at a sensor node is limited (e.g., due to battery size or energy-harvesting technology such as a solar cell), an energy saving mechanism is required and it can impact the wireless transmission performance significantly. A queueing model for sensor networks with energy conservation feature through sleep and wake-up mechanism was presented. A vacation queueing model was used to investigate the inter-relationship between the transmission performance and the energy consumption.

Power-Save Modes | WiMAX

IEEE 802.16e defines two new modes: the Sleep mode and the Idle mode in order to have:

• power-efficient MS operation;
• a more efficient handover.

Consequently, the normal operation mode that exists in 802.16-2004 is known as the Active mode.

1: Sleep Mode

In the Sleep mode state, the MS conducts pre-negotiated periods of absence from the serving BS air interface. The MS is unavailable to the serving BS (downlink and uplink) in these periods. The Sleep mode objectives are the following:

• minimise MS power usage;
• minimise the usage of the serving base station air interface resources.

In addition, the MS can scan other base stations to collect information to assist handover during the Sleep mode. Implementation of the Sleep mode is optional for the MS and mandatory for the BS.

For each MS in the Sleep mode, its BS keeps one or several contexts, each one related to a certain Sleep mode power saving class. The power saving class is a group of connections that have common demand properties. There are three types of power saving class, which differ by their parameter sets, procedures of activation/deactivation and policies of MS availability for data transmission. The MOB_SLP-REQ (SLeeP Request Message) (sent by a Sleep mode supporting MS) and the MOB_SLP-RSP (SLeeP Response Message) (sent by the BS) allow a request to be made for a definition and/or activation of certain Sleep mode power-save classes.

The unavailability interval of an MS is a time interval that does not overlap with any listening window of any active power saving class of this MS. During the unavailability interval the BS does not transmit to the MS, so the MS may power down or perform other activities that do not require communication with the BS, such as scanning neighbour BSs, associating with neighbour BSs, etc. During unavailability intervals for the MS, the BS may buffer (or it may drop) MAC SDUs addressed to unicast connections bound to the MS.

2: Idle Mode

The Idle mode is intended as a mechanism to allow the MS to become periodically available for downlink broadcast traffic messaging without registration at a specific BS as the MS traverses an air link environment populated by multiple BSs, typically over a large geographic area ^[2]. The Idle mode benefits the MSs by removing the active requirement for handovers and all Active mode normal operation requirements. By restricting MS activity to scanning at discrete intervals, the Idle mode allows the MS to conserve power and operational resources. The Idle mode also benefits the network and the BSs by eliminating air interface and network handover traffic from essentially inactive MSs while still providing a simple and fast method (paging) for alerting the MS about pending downlink traffic.

The BS are divided into logical groups called paging groups. The purpose of these groups is to offer a contiguous coverage region (see Figure 1) in which the MS does not need to transmit in the uplink yet can be paged in the downlink if there is traffic targeted at it. The paging groups have to be large enough so that most MSs will remain within the same paging group most of the time and small enough such that the paging overhead is reasonable. A BS may be a member of one or more paging groups.

Figure 1: Example of paging groups.

The MOB_PAG-ADV (BS broadcast PAGing) message is sent by the BS on the Broadcast CID or Idle mode multicast CID during the BS paging interval. This message indicates for a number of Idle mode supporting MSs a requirement to perform ranging to establish location and acknowledge a message or to enter the network. An MS will terminate the Idle mode and re-enter the network if it decodes a MOB_PAG-ADV message that contains the MS MAC address and an action code of 0b10 (Network Entry).

Idle mode initiation may begin after MS de-registration. During the Active mode normal operation with its serving BS, an MS may signal intent to begin the Idle mode by sending a DREG-REQ message with a De-Registration_Request_Code = 0 X 01, indicating a request for MS de-registration from a serving BS and initiation of the MS Idle mode. At MS Idle mode initiation, an MS may engage in cell selection to obtain a new preferred BS. A preferred BS is a neighbour BS that the MS evaluates and selects as the BS with the best air interface downlink properties.

The Idle mode can also be BS initiated: the serving BS includes a REQ duration TLV with an Action code = 0 × 05 in the DREG-CMD message, signalling for an MS to initiate an Idle mode request.

Fast BS Switching (FBSS) and Macro Diversity Handover (MDHO)

1: Diversity Set

There are several conditions that are required to the diversity BSs featured in FBSS and MDHO procedures. These conditions are listed below :

§  The BSs are synchronised based on a common time source and have synchronised frames.

§  The frames sent by the BSs from the diversity set arrive at the MS within the prefix interval, i.e. transmission delay

§  The BSs operate at same frequency channel.

§  The BSs are required to share or transfer MAC context. Such context includes all information MS and BS normally exchange during Network Entry, particularly the authentication state, so that an MS authenticated/registered with one of the BSs from the diversity set BSs is automatically authenticated/registered with other BSs from the same diversity set. The context also includes a set of service flows and corresponding mapping to connections associated with the MS, current authentication and encryption keys associated with the connections. There are also BS conditions specific to MDHO (see below).

An MS may scan the neighbour BSs and then select BSs that are suitable to be included in the diversity set. The MS reports the selected BSs and the diversity set update procedure is performed by the BS and the MS. After an MS or BS has initiated a diversity set update using MOB_MSHO/BSHO-REQ, the MS may cancel the diversity set update at any time. This cancellation is made through transmission of an MOB_HO-IND with proper parameters. The BS may reconfigure the diversity set list and retransmit the MOB_BSHO-RSP message to the MS.

In an MS diversity set, a member identifier, TEMP_BSID, is assigned to each BS in the diversity set.

2: Different Types of BS for a Given MS

Before getting into the details of make-before-break handover algorithms, FBSS and MDHO, the different types of BS for a given MS are summarized:

§  Serving BS. The serving BS is the BS with which the MS has most recently completed registration at the initial Network Entry or during a handover.

§  Neighbour BS. A neighbour BS is a BS (other than the serving BS) whose downlink transmission can be (relatively well) received by the MS.

§  Target BS. This is the BS that an MS intends to be registered with at the end of a handover.

§  Active BS. An active BS is informed of the MS capabilities, security parameters, service flows and full MAC context information. For a Macro Diversity HandOver (MDHO), the MS transmits/receives data to/from all active BSs in the diversity set.

§  Anchor BS. For MDHO or FBSS supporting MSs, this is a BS where the MS is registered, synchronised, performs ranging and monitors the downlink for control information (see Figure 1). For an FBSS supporting MS, this is the serving BS that is designated to transmit/receive data to/from the MS at a given frame. Hence, it can be verified that an anchor BS is a specific case of a serving BS. An MS is required continuously to monitor the signal strength of the BSs that are included in the diversity set. The MS selects one BS from its current diversity set to be the anchor BS and reports the selected anchor BS on the CQICH or MOB_MSHO-REQ message. The MSs and BSs may use the fast-feedback method to update the diversity set: when the MS has more than one BS in its diversity set, the MS transmits fast anchor BS selection information to the current anchor BS using the OFDMA fast-feedback channel.

Figure 1: Illustration of an anchor BS in a diversity set

3: FBSS (Fast BS Switching)

An FBSS handover begins with a decision for an MS to receive/transmit data from/to the anchor BS that may change within the diversity set. An FBSS handover can be triggered by either MOB_MSHO-REQ or MOB_BSHO-REQ messages.

When operating in FBSS, the MS only communicates with the anchor BS for uplink and downlink messages (management and traffic connections). The MS and BS maintain a list of BSs that are involved in FBSS with the MS. This is the FBSS diversity set. The MS scans the neighbour BSs and selects those that are suitable to be included in the diversity set. Among the BSs in the diversity set, an anchor BS is defined. An FBSS handover is a decision by an MS to receive or transmit data from a new anchor BS within the diversity set.

The MS continuously monitors the signal strength of the BSs of the diversity set and selects one of these BSs to be the anchor BS. Transition from one anchor BS to another, i.e. BS switching, is performed without exchange of explicit handover signalling messages. An important requirement of FBSS is that the data are simultaneously transmitted to all members of a diversity set of BSs that are able to serve the MS.

The FBSS supporting BSs broadcast the DCD message including the H_Add Threshold and H_Delete Threshold. These thresholds may be used by the FBSS-capable MS to determine if MOB_MSHO-REQ should be sent to request switching to another anchor BS or changing diversity set.

4: MDHO (Macro Diversity Handover)

An MDHO begins with a decision for an MS to transmit to and receive from multiple BSs at the same time. An MDHO can start with either MOB_MSHO-REQ or MOB_BSHO-REQ messages. When operating in an MDHO, the MS communicates with all BSs in the diversity set for uplink and downlink unicast traffic messages (see Figure 2). The use of this transmission diversity is not the same in the two different communications:

§  For a downlink MDHO two or more BSs provide synchronised transmission of MS downlink data such that diversity combining can be performed by the MS.

§  For an uplink MDHO, the transmission from an MS is received by multiple BSs such that selection diversity of the information received by multiple BSs can be performed.

Figure 2: Illustration of an MDHO operation mode

The BSs involved in an MDHO or equivalently a member of an MS MDHO diversity set must use the same set of CIDs for the connections that have been established with the MS. The same MAC/PHY PDUs should be sent by all the BSs involved in the MDHO to the MS.

The decision to update the diversity set and the process of anchor BS update begin with notifications by the MS (through the MOB_MSHOREQ message) or by the BS (through the MOB_BSHO-REQ message).

The Handover Process | WiMAX

The 802.16 standard states that the handover decision algorithm is beyond its scope. The WiMAX Forum documents do not select a handover algorithm either. Only the framework is defined. The MS, using its current information on the neighbour BS or after a request to obtain such information, evaluates its interest in a potential handover with a target BS. Once the handover decision is taken by either the serving BS or the MS, a notification is sent over the MOB_BSHO-REQ (BS Handover REQuest) or the MOB_MSHO-REQ (MS Handover REQuest) MAC management messages, depending on the handover decision maker: the BS or MS. The handover process steps are described in the following.

The handover process is made of five stages which are summarized in Figure 1. The HO process stages are described in the following sections

Figure 1: Illustration of handover process stages.

1: Cell Reselection

Cell reselection refers to the process of an MS scanning and/or association with one or more BS in order to determine their suitability, along with other performance considerations, as a handover target. The MS may use neighbour BS information acquired from a decoded MOB_NBR-ADV message or may make a request to schedule scanning intervals or sleep intervals to scan, and possibly range, the neighbour BS for the purpose of evaluating the MS interest in the handover to a potential target BS.

2: Handover Decision and Initiation

A handover begins with a decision for an MS to make a handover from a serving BS to a target BS. The decision may originate either at the MS or the serving BS. The handover decision results in a notification of MS intent to make a handover thruugh the MOB_MSHO-REQ (MS HO REQuest) message (handover decision by the MS) or the MOB_BSHO-REQ (BS HO REQuest) message (handover decision by the BS).

The BS may transmit a MOB_BSHO-REQ message when it wants to initiate a handover. This request may be recommended or mandatory. In the case where it is mandatory, at least one recommended BS must be present in the MOB_BSHO-REQ message. If mandatory, the MS responds with the MOB_HO-IND message, indicating commitment to the handover unless the MS is unable to make the handover to any of the recommended BSs in the MOB_ BSHO-REQ message, in which case the MS may respond with the MOB_HO-IND message with proper parameters indicating HO reject. An MS receiving the MOB_BSHO-REQ messsage may scan recommended neighbour BSs in this message.

In the case of an MS initiated handover, the BS transmits an MOB_BSHO-RSP message upon reception of the MOB_MSHO-REQ message.

3: Synchronisation to a Target BS Downlink

Synchronisation to a target BS downlink must be done. If the MS had previously received a MOB_NBR-ADV (MAC management) message including a target BSID, physical frequency, DCD and UCD, this process may be shortened. If the target BS had previously received hanndover notification from a serving BS over the backbone, then the target BS may allocate a non-contention-based initial ranging opportunity.

4: Ranging and Network Re-entry

The MS and the target BS must conduct handover ranging. Network re-entry proceeds from the initial ranging step in the Network Entry process: negotiate basic capabilities, PKM authentication phase, TEK establishment phase, registration (the BS may send an unsolicited REG-RSP message with updated capabilities information or skip the REG-RSP message when there is no TLV information to be updated) and the other following Network Entry optional steps (lP connectivity, etc.).

Network re-entry may be shortened by target BS possession of MS information obtained from the serving BS over the backbone network. Depending on the amount of that information, the target BS may decide to skip one or several of the Network Entry steps (Figure 2). Handover ranging can then be a simplified version of initial ranging. To notify an MS seeking handover of possible omission of re-entry process management messages during the current handover attempt (due to the availability of MS service and operational context information obtained over the backbone network), the target BS must place, in the RNG-RSP message, an HO Process Optimisation TLV indicating which re-entry management messages may be omitted. The MS completes the processing of all indicated messages before entering Normal Operation with the target BS.

Figure 2: Summary of network re-entry steps

Regardless of having received MS information from a serving BS, the target BS may request MS information from the backbone network.

5: Termination of MS Context

This is the final step of a handover. Termination of the MS context is defined as the serving BS termination of the context of all connections belonging to the MS and the discarding of the context associated with them, i.e. information in queues, ARQ state machine, counters. timers, header suppression information. etc. This is accomplished by sending the MOB__HO-IND message with the HO_IND_type value indicating a serving BS release.

6: Handover Cancellation

An MS may cancel HO at any time prior to expiration of the Resource_Retain_Time interval after transmission of the MOB_HO-IND message. Resource_Retain_Time is one of the parameters exchanged during the registration procedure (part of Network Entry). The standard indicates that Resource_Retain_Time is a multiple of 100 milliseconds and that 200 milliseconds is recommended as default.