180 likes | 382 Views
Policy Enabled Handoff Across Heterogeneous Wireless Networks. Helen W and Jochen Giese Computer Science, Berkeley. Outline. Abstract of the Paper Policy Enabled Handoff System - Case and Principles - Operating Environment - Policy Specification/Model
E N D
Policy Enabled Handoff AcrossHeterogeneous Wireless Networks Helen W and Jochen Giese Computer Science, Berkeley.
Outline Abstract of the Paper Policy Enabled Handoff System - Case and Principles - Operating Environment - Policy Specification/Model - Dealing with the Handoff Synchronization Problem Implementation of the Policy Enabled handoff System - Programming Model - Software Architecture Conclusions and Future Work
Abstract Integration of Heterogeneous networks Hand off issues in such networks (includes policy making) Software Architecture/Support for Heterogeneous Networks This paper presents a sample description of a heterogeneous wireless network architecture and focuses on a novel handoff mechanisms to achieve optimum performance. The heterogeneous infrastructure experimented was on four types of networks - Infrared LAN, WaveLAN, Metricomm, GSM Cellular.
Some terms/ideas of Interest Vertical Handoff- Intercellular Handoff- Handoff between cells of different networks. Features of Vertical Handoff- Vertical handoff is simple and embedded in the system.The mobile host always switches to the smallest coverage area. ************************************************************** Horizontal handoff- Intracellular Handoff- Handoff between cells of the same network. Feature of Horizontal Handoff- Horizontal handoff decision takes place based on received signal strength.
What was the Problem then……????? Firstly, The problem was sometimes frequent and unnecessary handoffs. Secondly, These handoff decisions did not consider system dynamics such as… Thirdly, the handoff synchronization problem.- The problem of overloading a network by mobile users simultaneously connecting at the same instant of time. So, Berkeley came up with some policy deciding when a handoff is necessary.
We here describe the factors that Berkeley thought that affect the handoff system • Non- Dynamic factors – • Bandwidth, latency • Power consumption • Charge Model • Dynamic factors – • Network conditions such as load, traffic…. • User conditions such as mobility…
Berkeley's Handoff Policy- Case and Principles Network technologies differ in terms of bandwidth, latency, power consumption and potentially their charge model. The issue was how to integrate these seamlessly. -Seamless ness was proposed to be achieved -Handoff latency was proposed to be low enough not to disrupt the running applications. The automation of switching from one network to another is based on the principle of user involvement with minimal user interaction. User involvement was required for the policy specification. Handoff decisions and operations are all done at the mobile host. Periodically, the mobile host collects current dynamic conditions, and consults with a pol- icy module on which is the best reachable network.
Policy Specification/Model Berkeley embedded its handoff decision policy to one equation The cost of using a network n is a function of several parameters such as the bandwidth it can offer(Bn), the power consumption of using the network access device (Pn) and the cost (Cn) using the network and is represented as... Costn = f(Bn, Pn, Cn) • For a given network…. • Bandwidth parameter is calculated periodically( Note that available BW • is inversely proportional to the number of users) • b) Power Consumption are fixed • c) Cost parameters are fixed
Policy Specification/Model (Contd…) Costn = f (Bn, Pn, Cn) Normalization is done….N(x)….. (Why…..) Costn = f (N(Bn), N(Pn), N(Cn)) The function is further expressed as Costn = wb*N(Bn), wp* N(Pn), wc*N(Cn)) Where wb, wp and wc are weight values and sum to 1. The setting of these values are determined by the user
These were Berkeley’s Specs for the simulated network The Simple Mobile IP Architecture follows…
Stability Period (To avoid frequent handoffs) Stability period is a defined as the waiting period before handoffs. The system calculates the cost function of each reachable network periodically and hands off to the better network if a network is consistently better for a stability period. Only if the network is consistently better than the current one in use for the stability period does the mobile host perform handoff. If a mobile user only transiently transfers to a better network, the gain from using the network may be diminished by the handover overhead and short usage duration.
Handoff Synchronization problem It is possible that several mobile hosts could discover the same better network and switch to them simultaneously causing it to increase its load dramatically. The synchronization problem can cause instability for all these mobile hosts and lead to poor performance. The problem is solved through randomized stability period. A random number is generated as the waiting period between handoffs.
Handoff Synchronization vs. Desynchronization The left diagram shows instability due to simultaneous handoffs whereas that instability is eliminated by incorporating stability period.
Implementation of the Policy Enabled Handoff System - Programming Model
Implementation of the Policy Enabled Handoff System – Software Architecture
Performance of the Berkeley Prototype Four networks were included in the prototype: 1) IBM Infrared LAN, 2) Lucent WaveLAN, 3) The Metricom Ricochet network, 4) and the GSM cellular. Performance was based on handoff latency. Handoff latency is the amount of time for a mobile host to handoff to a given network. Handoff latencies were measured from ten trials of manually triggered handoffs.
Future Work • Discovering Parameter Interdependency • This is important since varying one parameter can cause changes to the rest of them. It includes understanding and modeling the relationship between these system parameters. • 2) Redefining the policy model for various application levels • Users can specify priorities or different classes of QOS requirements for each application such as real time and non- real time. This can be easily incorporated into the policy model.