22
Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

  • View
    219

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Efficient Monitoring of QoS Parameters (EMQP)

Authors:Vadim DrabkinArie OrlovskyConstantine Elster

Instructors:Dr. Danny RazMr. Ran Wolff

Page 2: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 2

EMQP - Document Structure

Introduction Overview Terms and Conditions Proposed Solution Program Flow Implementation Issues Results Conclusion

Page 3: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 3

EMQP - Introduction

Today’s Internet delivers "best-effort" performance dictated by the very design of the Internet Protocol (IP). That is, traffic is processed as quickly as possible, but there is no guarantee as to timeliness or actual delivery. Consequently, important business applications packets such as real-time video or audio packets could be lost or to be delivered too late because of displacement by less important packets. This leads to the need for Quality of Service in order to provide a consistent predictable data delivery service for different kinds of packets. Quality of Service (QoS) is an umbrella term for a collection of technologies that allow network-aware applications to request and receive predictable service levels in terms of data throughput capacity (bandwidth), latency variations (jitter) and propagation latency.

Page 4: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 4

EMQP - Overview

The EMQP project comes to introduce an algorithm proposed by Dr. Danny Raz and Mr. Ran Wolff.

The EMQP aims to achieve an efficient monitoring of network traffic. – One of the main goals is discovering if the proposed

algorithm is more efficient than currently proposed techniques in the case of monitoring heavy traffic networks.

The EMQP project comes to test the proposed algorithm by means of network simulator, which has the monitoring capability of the proposed algorithm.

Page 5: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 5

EMQP – Terms and Conditions

Network Topology– a graph with nodes and edges

Node – a host connected to network Edge - hop-to-hop path between two nodes

Stream – A path between two nodes in a given network.

Latency – metric value of an edge – the cost of going from one of the edge’s end to the another

edge’s end. Latency threshold

– Local problem considered, if an edge’s latency is greater than its latency threshold.

Page 6: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 6

EMQP – Terms and Conditions (cont.)

Stream latency – The sum of latencies of stream edges

Threshold for stream latency – maximum allowed stream latency

Considered as a stream problem if greater than stream latency

Bandwidth Broker – The controlling host managing network

streams

Page 7: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 7

EMQP – Proposed Solution

The algorithm outline– Every edge checks if its latency exceeds its

latency threshold for every stream going through the edge.

If it does – trigger message to the next edge of the stream containing:

– Edge latency– Latency threshold– The stream identificator

Page 8: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 8

EMQP – Proposed Solution (cont.)

The algorithm outline (cont.)– If edge receives a message

Checks if received latency is greater than stream latency threshold

– If true – report about the stream problem to the bandwidth broker.

Otherwise: checks if the (received latency + edge latency) is greater than (received latency threshold + edge’s latency threshold)

– If true - send to the next stream’s edge: (received latency + edge) (received latency threshold + edge’s latency threshold) Stream identificator

Page 9: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 9

EMQP – Proposed Solution (cont.)

e2

e1

n1

n2

n3

latency = 2latency threshold = 1

latency = 4latency threshold = 6

Figure 3.

n2 triggers <route, 2, 1> message to n3.n3 gets the message, but does not forward any information to other edges, since

2 + 4 < 1 + 6

Page 10: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 10

EMQP – Program Flow

BriteBrite Parser Network

Model

Object

Stream Generator(BFS & Distribution)

NetworkModel

Containing streams

Object

EMQP

NetworkModel

Results Data

Object

Graph Generator

Text File

Page 11: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 11

EMQP – Implementation Issues

Message driven– Messages are stored in message queue

Delivered by nodes

– Messages are delivered to nodes Node answers whether it wants to deliver a new

message

Page 12: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 12

EMQP – Implementation Issues (cont.)

Page 13: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 13

Results 100 nodes100 streamsaverage fan out 4latency deviation 0.1average latency 0.93

Page 14: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 14

Results (cont.)

100 nodes100 streamsaverage fan out 4latency deviation 0.1average latency 0.97

Page 15: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 15

Results (cont.) 100 nodes100 streamsaverage fan out 4latency deviation 0.1average latency 1.0

Page 16: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 16

Results (cont.) 100 nodes100 streamsaverage fan out 4latency deviation 0.1average latency 1.025

Page 17: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 17

Results (cont.) 100 nodes100 streamsaverage fan out 4latency deviation 0.1average latency 1.05

Page 18: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 18

Results (cont.)

100 streams 200 streams

50 streams

Page 19: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 19

Results (cont.)

100 nodesVariable number of streams Average fan out 4Latency deviation 0.1Average latency 0.95

Page 20: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 20

Results (cont.)

50 streams

100 streams

200 streams

Page 21: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 21

Results (cont.)

100 nodes100 streams Variable average fan outLatency deviation 0.1Average latency 0.95

Page 22: Efficient Monitoring of QoS Parameters (EMQP) Authors: Vadim Drabkin Arie Orlovsky Constantine Elster Instructors: Dr. Danny Raz Mr. Ran Wolff

Vadim Drabkin, Constantine Elster, Arie Orlovsky, Technion I.I.T. 2002 22

Results - Conclusion

The EMQP algorithm has significantly better performance than the standard 1

EMQP algorithm has better performance than standard 2 when the possibility of edge problem decreases

– It leaves us the the main drawback of the EMQP algorithm, which appears when there is high possibility of message triggering  from several stream points, which in addition causes several messages sent to Bandwidth Broker addressed from the same stream.  In such cases we see that the standard 2 performs better despite of its simplicity

– It could require an extra work on finding how to eliminate the unnecessary traffic. One can think on a way of sending the messages to both directions and then eliminating the unnecessary message in their meeting point