Modeling Quality-Quantity based Communication Orr Srour under the supervision of Ishai Menache.

Slides:



Advertisements
Similar presentations
QoS Strategy in DiffServ aware MPLS environment Teerapat Sanguankotchakorn, D.Eng. Telecommunications Program, School of Advanced Technologies Asian Institute.
Advertisements

Game Theory and the Internet Flow and Congestion Control.
Hadi Goudarzi and Massoud Pedram
Brief introduction on Logistic Regression
01. Apr INF-3190: Congestion Control Congestion Control Foreleser: Carsten Griwodz
Congestion Control Algorithms
Channel Allocation Protocols. Dynamic Channel Allocation Parameters Station Model. –N independent stations, each acting as a Poisson Process for the purpose.
Traffic Shaping Why traffic shaping? Isochronous shaping
Abhay.K.Parekh and Robert G.Gallager Laboratory for Information and Decision Systems Massachusetts Institute of Technology IEEE INFOCOM 1992.
Playback-buffer Equalization For Streaming Media Using Stateless Transport Prioritization By Wai-tian Tan, Weidong Cui and John G. Apostolopoulos Presented.
A Transmission Control Scheme for Media Access in Sensor Networks Lee, dooyoung AN lab A.Woo, D.E. Culler Mobicom’01.
Sampling Distributions
DYNAMIC POWER ALLOCATION AND ROUTING FOR TIME-VARYING WIRELESS NETWORKS Michael J. Neely, Eytan Modiano and Charles E.Rohrs Presented by Ruogu Li Department.
Restricted Slow-Start for TCP William Allcock 1,2, Sanjay Hegde 3 and Rajkumar Kettimuthu 1,2 1 Argonne National Laboratory 2 The University of Chicago.
Generated Waypoint Efficiency: The efficiency considered here is defined as follows: As can be seen from the graph, for the obstruction radius values (200,
Measurements of Congestion Responsiveness of Windows Streaming Media (WSM) Presented By:- Ashish Gupta.
CPSC Topics in Multimedia Networking A Mechanism for Equitable Bandwidth Allocation under QoS and Budget Constraints D. Sivakumar IBM Almaden Research.
Differentiated Services. Service Differentiation in the Internet Different applications have varying bandwidth, delay, and reliability requirements How.
Lecture 9. Unconstrained Optimization Need to maximize a function f(x), where x is a scalar or a vector x = (x 1, x 2 ) f(x) = -x x 2 2 f(x) = -(x-a)
TCP Stability and Resource Allocation: Part I. References The Mathematics of Internet Congestion Control, Birkhauser, The web pages of –Kelly, Vinnicombe,
Performance Analysis and Monitoring Facilities in CPN Tools Tutorial CPN’05 October 25, 2005 Lisa Wells.
2001/10/25Sheng-Feng Ho1 Efficient and Scalable On- Demand Data Streaming Using UEP Codes Lihao Xu Washington University in St. Louis ACM Multimedia 2001.
Software Quality Control Methods. Introduction Quality control methods have received a world wide surge of interest within the past couple of decades.
High speed TCP’s. Why high-speed TCP? Suppose that the bottleneck bandwidth is 10Gbps and RTT = 200ms. Bandwidth delay product is packets (1500.
Ant Colonies As Logistic Processes Optimizers
A Scalable Network Resource Allocation Mechanism With Bounded Efficiency Loss IEEE Journal on Selected Areas in Communications, 2006 Johari, R., Tsitsiklis,
1 Spring Semester 2007, Dept. of Computer Science, Technion Internet Networking recitation #8 Explicit Congestion Notification (RFC 3168) Limited Transmit.
In this chapter we introduce the idea of a random variable as well as looking at its shape, center, and spread.
7/3/2015© 2007 Raymond P. Jefferis III1 Queuing Systems.
Ns Simulation Final presentation Stella Pantofel Igor Berman Michael Halperin
10th Workshop on Information Technologies and Systems 1 A Comparative Evaluation of Internet Pricing Schemes: Smart Market and Dynamic Capacity Contracting.
1 Incentive-Based Scheduling for Market-Like Computational Grids Lijuan Xiao, Yanmin Zhu, Member, IEEE, Lionel M. Ni, Fellow, IEEE, and Zhiwei Xu, Senior.
A Simulation Approach for Internet QoS Market Analysis Bruno Pereira Miguel Martins.
Switching Techniques Student: Blidaru Catalina Elena.
Distributed Quality-of-Service Routing of Best Constrained Shortest Paths. Abdelhamid MELLOUK, Said HOCEINI, Farid BAGUENINE, Mustapha CHEURFA Computers.
Dynamic and Decentralized Approaches for Optimal Allocation of Multiple Resources in Virtualized Data Centers Wei Chen, Samuel Hargrove, Heh Miao, Liang.
1 Performance Evaluation of Computer Networks: Part II Objectives r Simulation Modeling r Classification of Simulation Modeling r Discrete-Event Simulation.
جلسه دهم شبکه های کامپیوتری به نــــــــــــام خدا.
DELAYED CHAINING: A PRACTICAL P2P SOLUTION FOR VIDEO-ON-DEMAND Speaker : 童耀民 MA1G Authors: Paris, J.-F.Paris, J.-F. ; Amer, A. Computer.
CONGESTION CONTROL and RESOURCE ALLOCATION. Definition Resource Allocation : Process by which network elements try to meet the competing demands that.
Measures of Variability In addition to knowing where the center of the distribution is, it is often helpful to know the degree to which individual values.
Computer Networks with Internet Technology William Stallings
Aemen Lodhi (Georgia Tech) Amogh Dhamdhere (CAIDA)
© 2006 Cisco Systems, Inc. All rights reserved. Optimizing Converged Cisco Networks (ONT) Module 3: Introduction to IP QoS.
Competitive Queue Policies for Differentiated Services Seminar in Packet Networks1 Competitive Queue Policies for Differentiated Services William.
REECH ME: Regional Energy Efficient Cluster Heads based on Maximum Energy Routing Protocol Prepared by: Arslan Haider. 1.
Copyright 2008 Kenneth M. Chipps Ph.D. Controlling Flow Last Update
1 - CS7701 – Fall 2004 Review of: Detecting Network Intrusions via Sampling: A Game Theoretic Approach Paper by: – Murali Kodialam (Bell Labs) – T.V. Lakshman.
Information Theory for Mobile Ad-Hoc Networks (ITMANET): The FLoWS Project Competitive Scheduling in Wireless Networks with Correlated Channel State Ozan.
TCP-Cognizant Adaptive Forward Error Correction in Wireless Networks
6 December On Selfish Routing in Internet-like Environments paper by Lili Qiu, Yang Richard Yang, Yin Zhang, Scott Shenker presentation by Ed Spitznagel.
June 4, 2003EE384Y1 Demand Based Rate Allocation Arpita Ghosh and James Mammen {arpitag, EE 384Y Project 4 th June, 2003.
Measuring the Capacity of a Web Server USENIX Sympo. on Internet Tech. and Sys. ‘ Koo-Min Ahn.
Maximizing Lifetime per Unit Cost in Wireless Sensor Networks
Loss-Bounded Analysis for Differentiated Services. By Alexander Kesselman and Yishay Mansour Presented By Sharon Lubasz
OPERATING SYSTEMS CS 3530 Summer 2014 Systems and Models Chapter 03.
Lecture Network layer -- May Congestion control Algorithms.
Spring Computer Networks1 Congestion Control Sections 6.1 – 6.4 Outline Preliminaries Queuing Discipline Reacting to Congestion Avoiding Congestion.
Courtesy Piggybacking: Supporting Differentiated Services in Multihop Mobile Ad Hoc Networks Wei LiuXiang Chen Yuguang Fang WING Dept. of ECE University.
Lecture 4 Page 1 CS 111 Summer 2013 Scheduling CS 111 Operating Systems Peter Reiher.
OPERATING SYSTEMS CS 3502 Fall 2017
Corelite Architecture: Achieving Rated Weight Fairness
Internet Networking recitation #9
Empirically Characterizing the Buffer Behaviour of Real Devices
Switching Techniques In large networks there might be multiple paths linking sender and receiver. Information may be switched as it travels through various.
Lecture 19 – TCP Performance
Provision of Multimedia Services in based Networks
Scheduling Algorithms in Broad-Band Wireless Networks
Internet Networking recitation #10
Congestion Control, Quality of Service, & Internetworking
Presentation transcript:

Modeling Quality-Quantity based Communication Orr Srour under the supervision of Ishai Menache

Agenda The problem Design motivations Our model Simulations Summary & Conclusions

Regularly, price depends both on the quality and the quantity of the desired service. Quantity Quality $1 $6 0.5$ The Problem - QQ Quality Quantity

The Problem - QQ While so, networks still suffer from the old fashioned ”one price per all” attitude. Users usually pay a constant price for their bandwidth, regardless to its usage. Same Price (?)

The Problem - QQ Many QoS (Quality of Service) protocols have been recently developed in order to solve the problem above. We have chosen to use the “Differentiated Services” approach.

Differentiated Services This method simply attaches a quality tag to every packet that is being sent. Three bits in the TOS byte of the IPv4 header are used in order to determine the desired service type for the current packet Urgent Differ Normal

The Dynamic Nature Let’s look at two different events: the NBA final playoff game, and the US presidency election results announcement. High – Many many people Low – Not too many people Number of Users (Quantity) Low – Text based service High – Video based service Bandwidth About 0.5 hourAbout 1.5 hoursDuration of usage

The Dynamic Nature These events were predictable. What about unpredictable events such as news flashes ? This kind of events cause a dramatic change in the service requests characteristics.

The Dynamic Nature Our service supplier should be able to operate under our dynamic world – thus be flexible and rapidly adjustable.

Ordinary Basic Server Model One queue Incoming packets rate = I [Packets/Time] Packets handling rate = H [Packets/Time] When I>H overflow occurs. All the packets beyond the server’s capacity will be dropped (ignored) – the “tail- drop” phenomenon.

Our Server Model Our server will be divided into a filter and a buffer. The filter will be constructed out of several queues, each with its own length and packet- loss rate. We will use 10 queues in our filter. A packet in the 1 st (left most) queue has probability of 0.1 not to be dropped, while a packet in the 10 th queue will be handled for sure.

Our Server Model Every packet that reaches the server contains some information about where it should be placed (using Differentiated Services). The server is in charge of spreading the packets between the different queues in the filter. SPD (Service-based Packet Dropping) is the filtering scheme, used to drop packets according to the queue they are in.

Our Server Model Filter Ordinary Buffer Server

Our Server Model Every packet’s allocation decision can be represented as a random variable representing the desired queue. The type of this random variable will be a Gaussian. Its standard deviation is announced by the server. The user is free to choose the center of the Gaussian, and the total amount of packets he sends.

A Methodical Break Where did this idea come from? The roots of this project are the physical variables from the world of statistical mechanics. The work tries to construct the server under certain temperature (1/β) and users with chemical potential (μ). The users try to achieve equilibrium with their environment (the server), just like any physical system we know.

Some SPD examples Θ – The energy of the most probable packet. μ – The quantity descending coefficient (controls the total number of packets submitted to the server). β – The standard deviation of the Gaussian function. Θ = 0.5 β = 50 μ= 6 Θ = 0.5 β = μ= 6 Θ = 0.9 β = 6.6 μ= 3 Controlled by the serverControlled by the user

Our Server Model Filter: Buffer: Packet Arrival

Our Server Model Filter: Buffer: SPD:

Our Server Model Filter: Buffer: SPD: β

Our Server Model Filter: Buffer: SPD:

Our Server Model Filter: Buffer: SPD: Θ

Our Server Model Filter: Buffer: SPD:

Our Server Model Filter: Buffer: SPD: 90%

Our Server Model Filter: Buffer: SPD:

The User Model Let us model a simple user using 3 properties: Desired amount of information - the total amount of packets that should pass the server (- not being dropped) [packets]. User’s link capacity (bandwidth) - the maximal amount of information that the user can transmit per time – usually physically constrained [packets/time]. Maximal price - the maximal price the user is willing to pay in order to receive his desired energy [energy units].

Notations u - a single data packet. R – a single user. E(u) – u’s acceptance probability. N R (E(u)) – The total number of packets R submits with the same energy E(u). Price(E(u), N R (E(u))) – the price of the packet u. It is determined by its energy and number of similar packets R sends.

Flow of Events Declaration of the price function Construction of the queues Provisioning phase: Runtime phase: Announcement of β Announcement of Θ & μ

The User’s Utility Function Let us define the utility of a user as a variable which is scales independent : The optimal choice for the user will then become: [EU/EU] = [1] U user = E tot TotalCost E tot

The Server’s Utility Function b - the price of 1 unit of capacity space [EU]. Profits = (payment from users) – (capacity) * b[EU] At the provision phase: During the runtime phase:

Relations Demonstrations E vs. β & Θ E vs. β & μ

The Cost Function For methodic reasons, let us choose the following cost function: For low quantity, high price for high energies. For high quantity, high price for high energies & middle energies.

Average Packet Price Θ μ β=0.01β=0.51 β=2.01 β=7.01 β=9.51β=19.51

The Server’s Optimization Simulation Let us take a look on how the server chooses it’s optimal announcement. Let us consider the following energy consumption: Average of 500 units of energy

Total Number of Packets vs. β

Packets Allocations vs. β

Number of Users Playing the Game

Profits vs. β

What If a Sudden Increase in the Energy occurs? Average Energy 800 Number Of Users 1000 Average Energy 300 Number of Users 1300

What If a Sudden Increase in the Energy occurs? Under the right choice of β, the server can handle sudden changes in its demands. The right choice here means that the total number of packets the server needs to handle does not vary too much, and thus no physical constraint will affect the server’s operations.

One Day Game After showing how the server chooses β, let us introduce a simple simulation that will try to estimate the network traffic for "one day”. Different hours of the day will gain different services desires. Let us choose the queues of the server to be 10,000 packet places long each.

Game Details 1.The provider is free to choose it’s price function once, and then change β during the game. 2.The provider has no energy limits (no maximal capacitance). 3.Every user has a constant desired energy - the energy he wishes to attain. 4.Every user has a constant maximal price he is willing to pay for his desired energy. 5.When the minimal price for the desired energy is more than the user's maximal price, the user stops playing (his energy equals to 0). 6.The user has a constant maximal capacity of packets he can handle (this is the maximal value of N per user). 7.The user tries to find the minimal price for his desired energy, and once it is found, he tries to find the highest energy for this price, so that the minimal energy he will receive is his desired energy.

Number of Playing Users Vs. Time

Maximal Price of 1 Packet vs. Time

The Average Total Energy Consumption

β During the Day

Number of Participating Users

Server’s Utility vs. Time

Packets Allocation vs. Time

Profits vs. Time

Summary and Conclusions In real life, the most general model should allow one user to send multiple requests for services at the same time. We observed high correlation between the different variables of the problem. Slight change in the average energy for example, as shown before, can dramatically change the number of users participating the game.

Summary and Conclusions In spite of the high correlated variables, we were able to answer our server needs: Under the right design, a server based on our model can offer multiple services, based upon Quality & Quantity. We showed that our server is able to handle sudden changes in the needs of its users.

Finally, The End Thank you for listening.