Table of Contents
21-08-2015 Telco 14 - 14:
FS MK SP TD - CM
IPR Discussion - literature search
Mk starts with slide 12
paper 1:
CM: we detail on history and broader mobility in geo. and temporal.
CM: where we innovate is not introducing learning but define details of parameters for learning: combination of 3 ideas (threadhold, cluster, mobilty pattern) in learning.
Ref 2,3,5 to be worked in details
13-08-2015 Telco 14 - 14:40
ZD CM / MK TD
IPR idea discussion
MK: discuss modules & component in the scenarios MK: Discuss candidate cluster
CM: Are we discuss mcast in the scenario? MK: The candidate cluster to be mcast to
MK: Build cluster by calculate Prob. of an AP to be PoA and compare with Threadhold ZD: We need to propose how to calculate Threadhold
MK: Discuss different triggers
MK: discuss settings 1
MK: When to start and stop Multicasting? we could stick to make-before-brake strategy
MK: discuss setting 2
ZD: agreed.
MK: discuss setting 3
ZD: we could later publish paper
ZD: for the patent we don'nt need all details, but specific. Give example of thread hold e.g,. For IPR needs details how we implement. How TH is selected, how probability is calculated but we don't need to check over that.
MK: details on cluster, probability thread hold.
ZD: you propose more complex mobility management: user, learning… we should match
MK: should we make it more generic or go to sub set of it.
ZD: go to sub set and next call we discuss multicast in details
MK: we work on details on cluster, thread hold, multicast
ZD: look for related works.
MK: if there are related works then we give up.
CM: you take a look at threadhold. Find the difference to what has been proposed.
MK: We start work on Thread hold function.
CM: Figure out the parameters for the functions.
ZD: May be more information about user, mobile node, how to collect the information. May be there are concretes in the environment that are important
MK: find grounding for the feasibility of the solution.
MK: take away: concentrate on cluster, thread hold, multicasting. Work on threadhold function. Look if the needed information for the solution is feasible.
CM: How is the MD-SAL implementation.
MK: there is some delay,
CM: Could you send AD-SAL implementation.
MK: we have the first version or we send the improved version next tuesday.
CM: want to learn about ODL and need some examples.
MK: we send the code when ready.
CM: How is the Survey paper?
MK: is making big changes based on HW comments. we will send when ready.
06-08-2015 Telco
- M: Agenda from MoM at F2F meeting
- M: Discuss paper Learning mobility behavior
- M: should we leave out architecture from paper scope?
- Z: let the architecture there with more details.
- M: we add details about learning.
- M: 1. implementation on static topology. 2. implementation with dynamic topology.
- Z: possible also submission to ICIN (less scientific).
- C: Suggest using information from SDN platform in the learning.
- M: we also learn over QoE of mobile user. User may select different PoA.
- C: Suggest support for different topologies.
- M: the algorithm learns over unknown environment. With real environment.
- Z: The paper can be more general. Show that learning can be useful for SDN. Type of information can be used for learning: flows statistic, user mobility prediction. Show that learning can improve mobility management.
- M: So this implementation is a sample of more general learning approach. We agreed.
- M: Book chapter proposal
- M: discuss SDN chapter proposal
- Z:
- M: cloud federation book chapter, activity from DAI students
- M: Event Reactive - behavioral proactive paper
- M: We include agent in the solution. Start research feasibility of integrating agent in SDN. Pursue IPR ideas in this context.
- Z: Focus on this later.
- M: MD-SAL + Survey Paper
- M: Implementation ready in 1 week. Then implement learning with MD-SAL and compare 3 implementation.
- Z: Speed up the development. Top priority for us to compare the approaches.
- M: in 3 week to discuss some tangibles.
- M: discuss C's comment on why not put densification in mobility management.
- C: It's like there is no link between the two topics. If we want to go in details of physical layer.
- M: We not really want to include SDR,
- C: State the focus of the paper but provide links to the topic in the right balance.
- M: We think over if we separate or merge those topics.
- Z: We workout the best of the paper.
- M: we provide new version before the next meeting.
- M: IPR ideas
- M: what is the potential idea to start?
- Z: multicasting. compose use-cases slides. we don't need result, just the idea.
- M: We work out how multicasting should be.
- Z: We continue with your use-case slide.
- M: provide usecase slides (2-3) till next week.
- Z: we look at the bookchapter and send feedbacks.
28-05-2015
- Restful implementation:
- Zoran: Learning Mobility pattern on high way setting. Sequential dis/connect along an array of switches. Turn the result to a paper.
- AD-SAL Reactive implementation:
- Zoran: We may target big conferences: ICC
- Zoran: More complex test scenarios, concurrent host mobility, many switches
- Z: what is AD/MD-SAL?
- M: Explain
- Z: how do you move host in mininet?
- M: In mininet, disconnect host from old switch and connect it with new one. Modification of mininet mobility script.
- C: How exactly mobility script work? how to do the ping?
- M: Node move then ping. A host to host ping can be observed from host terminal.
- MD-SAL implementation
- Z: Discuss limitation of ODL, Floodlight regarding synchronization of topology changes. More detailed in later discussion.
- IPR:
- Z: Target small innovation, add simple module to the controller. Whatever idea we believe to be novel.
- M: priority for IPR task?
- Z: IPR is most important. Identify small innovation and turn to patent. E.g, learning mobility pattern, connect the learning with mobility.
- Roadmap
- Z: Deliver papers and patent are most important.
- Z: next call in 3 week then plan the F2F meeting.
