The MAC Metric

MAClogo.gif
 


 
An evaluation procedure, namely a metric, quantitatively measures the suitability of available connectors.

The MAC metric exploits both static context information, e.g., average power consumption and interface nominal bandwidth, and dynamic context data, e.g., mobility state of clients and peer connectors. Those context data are effectively monitored and taken into account to provide the best tradeoff between connection durability and quality for applications running at a client node.

The MAC metric is based on the home-office-home mobility pattern assumption: users are expected to move by alternating movement and still phases. In many everyday-life situations, user requirements and surrounding environment vary in relation to user mobility state, either still or motion. For instance, a client in the still phase is expected to find infrastructure connectors available, with good quality and short/medium coverage range, e.g., IEEE 802.11 APs. In addition, while still, clients tend to require high channel reliability. On the opposite, the motion phase assumes only the presence of infrastructure connectors with large radio coverage, e.g., UMTS base stations. Moreover, users in motion usually decrease their network performance requirements, e.g., because they only perform background network activities such as email download. For these motivations we claim the need for middleware solutions that dynamically change their evaluation process also depending on user mobility state.

We have thoroughly followed that design guideline in MAC.

  • When a user is fixed (still state), MAC assumes that there is high probability of the availability of limited/medium coverage range infrastructure-based connectors (Bluetooth and IEEE 802.11) or fixed peer-based connectors (such as the nearby desktop PC of a user currently working on her PDA). In both cases, the primary objective of channel durability can be easily achieved: in fact, the above connector types usually provide sufficient QoS, are reliable, and inexpensive. In these cases, MAC should simply select the connector of the above types that maximizes bandwidth while minimizing power consumption. In any case connectors of these types should be preferred to both GPRS/UMTS base stations (often with non-negligible costs) and mobile peers (unreliable).
  • When a user is in the motion state, it is suitable to first choose infrastructure-based connectors with medium/large radio coverage, thus increasing the probability of connection durability. When infrastructure connectors are not available, MAC also considers mobile peer connectors with joint mobility with regards to requesting clients. If even joint connectors are not available, MAC additionally explores the opportunities offered by transient connectors with at least medium range wireless coverage, e.g., IEEE 802.11 mobile peers. In fact, transient peers with limited coverage range should be avoided, if possible, to reduce frequent handovers. The below table summarizes the above considerations, depending on the motion/still client state.
Based on these guidelines, for each considered client, MAC selects a preferred type of connector and limits its evaluation process to the set of eligible connectors of that type, if any available. MAC quantitatively evaluates every eligible connector by determining a ConnectorValue in the [0, 1] range (0=worse choice, 1=best choice) for each of them. To that purpose, it exploits the evaluation function:


ConnectorValue = EnduranceValue + QualityValue

(1)

where EnduranceValue estimates the expected connector durability and QualityValue its expected quality in terms of bandwidth and energy consumption.

The MAC metric depends on client motion/still state.

Client
mobility state

Wireless
interface

Connector
type

Still

Bluetooth ≈ 802.11 >> UMTS

fixed connector >> joint peer

Motion

802.11 ≈ UMTS >> Bluetooth

infrastructure ≈ joint peer >> transient peer

(≈ equivalent, > better, >> much better)

The evaluation of each term dynamically changes depending on the fact that the evaluated connector is either fixed or mobile. For a fixed connector:


EnduranceValue = 2
QualityValue = 4

(2)

while for a mobile connector:


EnduranceValue = 6
QualityValue = 8

(3)

CMob and Joint are values in the [0, 1] range and model the MAC estimations for, respectively, client and connector mobility. In particular, Joint represents the probability of relative mobility of the connector with regards to the associated client. MAC determines the original CMob and Joint indicators as detailed in Section 6.3. The Range parameter (in the [0, 1] interval) concisely models the radio coverage of a connector and only depends on its associated interface. For instance, MAC assigns a Range of 0.7 to IEEE 802.11 AP connectors while Range for Bluetooth peer connectors is 0.3. QualityValue functions consider the Bandwidth and Energy parameters (always in the [0, 1] interval), which model the ratio between, respectively, available/required bandwidth and energy. α and β (α, β ≥ 0 and α + β ≤ 1) concisely model user-specified application-level requirements, thus adapting the relative relevance between bandwidth and energy preferences.

Let us rapidly note that the equation pair for mobile connectors does not include CMob. In fact, based on the home-office-home assumption, MAC considers mobile peers only if there are not fixed connectors in visibility; with no fixed connectors in visibility, it is impossible to estimate CMob (see MAC Components). Finally, reliability is not included in the above evaluating functions because MAC automatically excludes connectors not compliant with user reliability requirements from the set of eligible connectors.

 
9-nov-09