Traffic and Cost Generator

Slides:



Advertisements
Similar presentations
NETWORK LAYER (1) T.Najah AlSubaie Kingdom of Saudi Arabia Prince Norah bint Abdul Rahman University College of Computer Since and Information System NET331.
Advertisements

Consistency and Replication Chapter 7 Part II Replica Management & Consistency Protocols.
1 Accessing nearby copies of replicated objects Greg Plaxton, Rajmohan Rajaraman, Andrea Richa SPAA 1997.
Beneficial Caching in Mobile Ad Hoc Networks Bin Tang, Samir Das, Himanshu Gupta Computer Science Department Stony Brook University.
Chapter 4 Network Layer slides are modified from J. Kurose & K. Ross CPE 400 / 600 Computer Communication Networks Lecture 14.
TDC 461: Wide Area Network Services DePaul University 22 May, 2001.
TCP/IP Protocol Suite1 Copyright © The McGraw-Hill Companies, Inc. Permission required for reproduction or display. Chapter 11 Unicast Routing Protocols.
MODULE IV SWITCHED WAN.
The Network Layer. Network Projects Must utilize sockets programming –Client and Server –Any platform Please submit one page proposal Can work individually.
Chapter 7 WANs and Remote Connectivity. WAN Essentials A WAN traverses a large geographic area A WAN link is a connection from one site to another and.
Routing Information Protocol (RIP). Intra-and Interdomain Routing An internet is divided into autonomous systems. An autonomous system (AS) is a group.
Outlines Received due 13 March 24 %. Homework n Review for Midterm on 1 March 2000 n Research Paper n Readings [11] “ATM Systems: What’s Next?” [12] “Long.
Network Layer4-1 Chapter 4: Network Layer Chapter goals: r understand principles behind network layer services: m network layer service models m forwarding.
4: Network Layer4-1 Schedule Today: r Finish Ch3 r Collect 1 st Project r See projects run r Start Ch4 Soon: r HW5 due Monday r Last chance for Qs r First.
Packet switching network Data is divided into packets. Transfer of information as payload in data packets Packets undergo random delays & possible loss.
Lecture # 03 Switching Course Instructor: Engr. Sana Ziafat.
3-location Data Network Design 3 locations separated by 200 km among pairs. Give the new populations below 296 users, design the data network.
1 TCOM 5143 Lecture 9 Traffic and cost generators.
WAN – Packet and Cell-switched Networks
11 ROUTING IP Chapter 3. Chapter 3: ROUTING IP2 CHAPTER INTRODUCTION  Understand the function of a router.  Understand the structure of a routing table.
RIP Routing Protocol. 2 Routing Recall: There are two parts to routing IP packets: 1. How to pass a packet from an input interface to the output interface.
Network Layer (2). Review Physical layer: move bits between physically connected stations Data link layer: move frames between physically connected stations.
Data Communication Networks Lec 13 and 14. Network Core- Packet Switching.
Internet Service Providers and types of internet connections
INTRODUCTION NETWORKING CONCEPTS AND ADMINISTRATION CSIS 3723
Chapter 2 PHYSICAL LAYER.
Chapter 3 Part 3 Switching and Bridging
Congestion Control in Data Networks and Internets
Lab A: Planning an Installation
Circuit Switching and Telephone Network
Chapter 8 Switching Copyright © The McGraw-Hill Companies, Inc. Permission required for reproduction or display.
Scaling the Network: The Internet Protocol
URL: Chapter 8 Switching Tel: (03) Ext: URL:
Routing Information Protocol (RIP)
Network Configurations
Wide Area Network.
Network Layer Goals: Overview:
Chapter 6 – Routing.
06- Transport Layer Transport Layer.
LAN Risanuri Hidayat.
CS4470 Computer Networking Protocols
Chapter 3 Switching.
Switching Techniques In large networks there might be multiple paths linking sender and receiver. Information may be switched as it travels through various.
Chapter 3 Part 3 Switching and Bridging
Wednesday, September 19, 2018 What Is the Internet?
Accessing nearby copies of replicated objects
Routing.
Design Unit 26 Design a small or home office network
Dr. John P. Abraham Professor UTPA
Lecture 5: TELEPHONE NETWORK
Chapter 20 Network Layer: Internet Protocol
Dr. John P. Abraham Professor UTRGV, EDINBURG, TX
70-293: MCSE Guide to Planning a Microsoft Windows Server 2003 Network, Enhanced Chapter 4: Planning and Configuring Routing and Switching.
Connectivity methods. Connectivity methods Exam Answer WAN/LAN  Each office will have its own Local Area Network or LAN.  But the company also wants.
Data Communication Networks
Network Layer I have learned from life no matter how far you go
Dr. John P. Abraham Professor UTPA
Capacity Unit 26 Design a small or home office network
3-location Data Network Design
Chapter 3 Part 3 Switching and Bridging
Scaling the Network: The Internet Protocol
COMP/ELEC 429/556 Fall 2017 Homework #2
CSE 550 Computer Network Design
Circuit Switched Network
Resource Negotiation, Pricing and QoS
Routing.
Switching A network is a set of connected devices
EECS 122: Introduction to Computer Networks Packet Scheduling and QoS
Private Networks.
Switching.
Presentation transcript:

Traffic and Cost Generator Real network design is 90% of file preparation and data collection. Some data need to be generated: Future growth prediction Large traffic demands (requirements, end-to-end traffic) data. Only limited info available. Traffic generator can fill in the gap. 11/19/2018 C. Edward Chow

Delite Network File Format Goal: simple extensible. 11/19/2018 C. Edward Chow

Tariff, Equipment, Param Format 11/19/2018 C. Edward Chow

Cluster Site Notation Here the PARENT column deal with the homing of this site. 11/19/2018 C. Edward Chow

Sites Population and Level 11/19/2018 C. Edward Chow

Traffic Units for Various Networks 11/19/2018 C. Edward Chow

Traffic Generation Assume Uniform Traffic. Traf(i,j)=C. Traffic from node i to node j is C. Assume Random Traffic . Traf(i,j) is some distribution between a minimum and a maximum value. What distribution is the following code generated? 11/19/2018 C. Edward Chow

Random Mux Circuit Traffic Generation D56 56kbps links Nreq circuits will be generated. 11/19/2018 C. Edward Chow

More Realistic Traffic Formula 1 11/19/2018 C. Edward Chow

Add Offset and Scale Factor A is scale factor Population offset (say 0.05) avoid zeros; Distance Offset. 11/19/2018 C. Edward Chow

Traffic Normalization Example 4.1: 50 sites linked by 85 E1 lines. The average of hops is 2.75 and the links have an average utilization of 0.55. What value of a should be chosen to generate the traffic? Solution: Total carried traffic on the network = Let T= We choose a =69,632,000/T 11/19/2018 C. Edward Chow

Row Normalization With row indicating source, column indicating destination in the traffic matrix. The traffic from node i to other nodes is We can then define This allows us to match the synthetic traffic to the observed traffic. 11/19/2018 C. Edward Chow

Row and Column Normalization 11/19/2018 C. Edward Chow

Example: 5 node network Site data and observed Traffic in and out of a node. 11/19/2018 C. Edward Chow

1st round with Pop_Power=1 and Dist_Power=1 11/19/2018 C. Edward Chow

1st round Row and Column Scale Factors and Modified Traffic 11/19/2018 C. Edward Chow

2nd Round Traffic Matrix 11/19/2018 C. Edward Chow

2nd Round Row and Column Scale Factors 11/19/2018 C. Edward Chow

5 Iteration later 11/19/2018 C. Edward Chow

Considering Level (Traffic Type of Nodes) Level Matrix (uneven traffic btw Level 1 and Level 2) Formula 11/19/2018 C. Edward Chow

Final Traffic Formula Consider the random traffic, 0<=rf: random fraction<=1 11/19/2018 C. Edward Chow

Traffic Generators and Sensitivity Analysis Often useful to generate traffic suites than a single set of traffic. It can be used to study how network responds to change. Sometimes, there is a requirement that network not be re-designed for a period of time. 11/19/2018 C. Edward Chow

A Case Study: “Green Field” Redesign of a Network Green field mean “unconstrained to reuse the network already in place. 7 nodes in Squareworld. Distance Matrix: 11/19/2018 C. Edward Chow

Parameter for Terminal Traffic N1-N5 each has 100 users, N6-N7 each has 200 users. Level one has 600 users. 11/19/2018 C. Edward Chow

Modeling User Session Traffic Matrix (by observing control point data) Here N4 and N7 have host computers. 11/19/2018 C. Edward Chow

Voice vs. Email Cost Fixed cost of email vs. time dependent cost of voice 11/19/2018 C. Edward Chow

Internal Email Traffic Parameter Total population=900. 8 2000bytes email in busy hour. Why only normalize on ROW? 11/19/2018 C. Edward Chow

Internal Email Traffic Matrix Generated by Traffic generator using the parameters in previous page. Note that the total email traffic includes those that send to destination in the same site. 11/19/2018 C. Edward Chow

Internal Email Actual Traffic Remove the diagonal entries leave the inter-site traffic. 11/19/2018 C. Edward Chow

Internal Web Page Traffic 25% web fetches are internal. 23 pages/hour/user Each page access results 5 128Byte datagrams outbound and 5 128 Byte datagrams inbound, 3500 byte HTML and related files inbound. Outbound traffic/user=0.25*23*5*128*8/3600=8.177bps Inbound traffic/user=8.177+0.25*23*3500*8/3600=52.9bps. TRAFIN and TRAFOUT only consider traffic initiated from the site. 11/19/2018 C. Edward Chow

Internal Web Page Traffic Parameter The sum of row traffic = TRAFOUT 11/19/2018 C. Edward Chow

Internal Web Page Traffic Out 92*4+180*2=730? Vs 818 This is due to 818*(1/9)=92 of the traffic is intra-site. The other 818*(8/9)=727 is the actual TRAFOUT that spread to all other 6 nodes. 183*5+362=1277? Vs 1636 due to Arbitrary Dist_offset? 11/19/2018 C. Edward Chow

Internal Web Page Traffic Inbound The ratio between the inbound traffic and outbound traffic = 52.90/8.177=6.46875. The inbound traffic matrix=6.46875* Ttr where Ttr is the transpose of matrix T. 11/19/2018 C. Edward Chow

External Web Page Traffic 75% of web traffic to external sites. 23 pages/hour/user Each page access results 5 128Byte datagrams outbound and 5 128 Byte datagrams inbound, 3500 byte HTML and related files inbound. Outbound traffic/user=0.75*23*5*128*8/3600=24.533bps Inbound traffic/user=24.533+0.75*23*3500*8/3600=158.70bps. Typo in text: “8.17 plus” should be “24.533 plus” 11/19/2018 C. Edward Chow

Partition DB Site Traffic Total DB=200GB, 80 GB at N1, 60 GB at N4, 60GB at N7. 900 users. Per user has 15 queries/hour, 300B request(out), 4000B data(in). 3 updates/hour, 8000B(out), 1000B(in). Query traffic to server: 15*300*8/3600=10bps. Query traffic from server: 15*4000*8/3600=133.333bps. Update traffic to server: 3*8000*8/3600=53.333bps. Update traffic from server: 3*1000*8/3600=6.666 bps. 100*(10+53.33)=6333 100*(133.3+6.6)=14000 63.33*900*0.4=22,799 140*900*0.4=50,400 N1 has 80GB, 40% of Total DB. 11/19/2018 C. Edward Chow

Partition DB Traffic Inbound 40% of N2 6333 outbound DB traffic go to N1=0.4*6333=2533. 0.3*6333=1900 to N4 and 1900 to N7. 11/19/2018 C. Edward Chow

Partition DB Traffic Outbound The ratio between the inbound traffic and outbound traffic = 14000/6333. The inbound traffic matrix=14000/6333* Ttr where Ttr is the transpose of matrix T. 11/19/2018 C. Edward Chow

Replicated DB Client-Server Query DB are replicated at site N1, N4, and N7. Assume static allocation of clients to servers. N5N1, N3N4, N2,N6N7. N5 has 100 pop*10bps/pop=1000 bps query outbound; 100*133.33=13,333 bps query inbound. 11/19/2018 C. Edward Chow

Replicated DB Client-Server Update DB are replicated at site N1, N4, and N7. Assume static allocation of clients to servers. N5N1, N3N4, N2,N6N7. N5 has 100 pop*53 1/3 bps/pop=5333 bps update outbound; 100*6 2/3=666 bps update inbound. 11/19/2018 C. Edward Chow

Replicated DB Update Server-Server Same 53 1/3 bps update request per user and 6 2/3 bps update response per user. N4 has update requests from N3 and N4 (200 users)= 200*53 1/3=10,666. Those will relay to N1 and N7. N4 also responds to N1’s update (including N1 and N5, 200 users)=200*6 2/3=1333. Therefore it is 1333+10,666 to N1. N1 has update requests from N1 and N5, 200 usres=200*53 1/3=10,666. N1 also responds to N7’s update (including N2,N6, N7, 500 users)=500* 6 2/3=3333. Therefore it is 10,666+3000 to N7. 11/19/2018 C. Edward Chow

Usage-Sensitive Voice Tariff In most data networks, the cost of bandwidth is the largest expense item. A router $10,000 is amortized to $300/month. It can terminate links that cost $20,000/month. 11/19/2018 C. Edward Chow

Usage-InSensitive Voice Tariff Banded Wide Area Telephone Service (WATS). Banded WATS (4 hours/day) means a service that allow up to 4 hours of call per day to locations within a certain distance, say 250 miles. 11/19/2018 C. Edward Chow

ISDN 11/19/2018 C. Edward Chow

Usage-InSensitive DataTariff (UK US tariff is much cheaper. 11/19/2018 C. Edward Chow

Tariff Taxonomy Four type of links: Fixed virtual circuit (leased link for voice and data) installation cost= 10~20 monthly rent Dialed virtual circuit. (setup on demand; limited speed choices) Fixed pipes: accept bit at certain rate and make the best effort to deliver it, e.g., X.25, frame relay, Switched Multi-megabit Data Service (SMDS). Two rates: peak rate (circuit rate), say 64kbps, and committed information rate (CIR), say 16kbps, guaranteed. Dialed pipes. Setup on demand; usage-sensitive charge. List of possible fees: Access fees (for maintaining the connections) Setup fees Teardown fees Usage fees. Depending on Channel capacity, CIR, Distance, Time of Day, National and administrative borders 11/19/2018 C. Edward Chow

Cost Model Distance-based Costing There are anomalies just like airfare due to competition Linear Distance-Based Costing=a+b*d where a=fixed cost; b=variable cost; d=distance a and b can be derived from samples in the tariff table using least-square curve fitting. For example, at UK cost=$757.09+$2.40/km*d at US cost=$605.00+$0.49/mile*d Piecewise-Linear Distance-Based Costing Piecewise-Constant Distance-Based Costing 11/19/2018 C. Edward Chow

Piecewise-Linear Distance-Based Costing Many tariffs are published in piecewise-linear format. T1 cost in US/Mexico 56Kbps cost in US/Canada 11/19/2018 C. Edward Chow

Piecewise-Constant Distance-Based Costing Banded charges in Japan for D64 link. 11/19/2018 C. Edward Chow

Tariff Related Issues No direct circuit (unusal connect)Tariff not filed. Low cost tariff  when you go to provision the circuit, there may not have facilities or take long time to get on. In US, there is a notion of Local Access and Transport Area (LATA). Within LATA is much cheaper than in end-points at different LATA. There may be multiple carriers and require optimization. 11/19/2018 C. Edward Chow

Cost Generators Implemented in cost-gen.c Cost generator 1: use C(i,j,k)=Fk+d*DCk site I to site j with type k link. DC variable cost. F: Fixed cost. Cost generator 2: use Cost generator 3: take into consideration different nations. Cost generator 4: involve two countries, two half-circuit costs Cost generator 5: override case 11/19/2018 C. Edward Chow