more than 50 % of traffic was generated by the Audio-galaxy music files transfers. ⢠The packeteer solution will help to identify and to shape the traffic accordingly ...
Quality of Service Provision Assessment for Campus Network Dr., Eur. Ing. Eduard Babulak B.Sc., M.Sc., Ph.D., C.Eng., SMIEEE, MIEE, MBCS, MACM Researcher - Senior Lecturer School of Computing Staffordshire University Stafford ST18 0DG UNITED KINGDOM Web: http://www.soc.staffs.ac.uk/~eb12 September 10-11, 2003
1 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Outline of Presentation • • • • •
Introduction CISCO Vision Maximizing Productivity Research Aims and Objectives QoS Definition – QoS View Points – QoS Layered Model
• • • •
Users’ Applications Features Enhanced Disconfirmation Model Determinants of Performance Quality Case Study: Staffordshire University Network – Network Measurements and IT Survey – Discussion of Results
• Conclusions – Question & Answers 2 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Cisco’s Vision: “Extending the Productivity Zone” 100% 70% 50%
30% 50% 70% 100% SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
3 by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Maximizing Productivity Throughout the Workday At Work • Conference rooms • Cafeteria • Temporary cubicle
PSTN
On the Road Internet/Intranet
• Hotels • Airports • Convention centers • Coffee shops • Customer sites
At Home • Telecommuting • Home office 4 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Cisco Mobile Office—On the Road A global program promoting secure, high-speed access in public venues for mobile professionals—delivered by industry-leading public venues, service providers, and integrator partners Planes, Trains
Customer Sites
Cisco Mobile Office— On the Road Other Venues
Airports
Convention Centers SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
Hotels by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
5 5
Final Thoughts • Investments in information technology have had significant impacts on organizational productivity and have set the foundation for the networked virtual organization • But the world is changing…and the work environment needs to change to better meet the challenges of the 21st century • Successful organizations will identify ways to extend their employees productivity zones by giving them extended access to the tools necessary to do their jobs 6 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Current & Future Challenges: Server
• QoS Provision
Real Audio
– Highly Interactive Applications Workstation
• Support Interoperability – Heterogeneous Environments – Open Networking
Camera
Home Video
Network
• Support Integration – Wide range of End devices
Video Title
• CISCO ATM Switch, IP Router with standard interfaces
Disc
– Internet Openness • Attach any network using IP
7 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Introduction • Current methods to assess the QoS provision – What performance is now in the future optimal? – What priority should be assigned to improve performance? – How to assess the user’ perception and network usage? • Identify the University Network Infrastructure – Campuses, Schools and Departments and their mutual links
• Identify basic categories of the University users – Understand the University Network initial Design, etc.,
• Be aware of current and future challenges – – – – –
Network Critical Applications (i.e., payroll, enrolment) Increased number of users and their mobility; More dynamics in traffic and applications; E-commerce and Distance Learning; Students using pattern of IT network • Entertainment and Music files download 8
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Research Aim & Objectives • Merger of Engineering and Business Perspectives on QoS Provision in Light of User’s Perception • Enhanced Disconfirmation Model • Network Traffic Measurements • IT Survey
• Suggest Performance Quality Improvements
•
• Determinants of Performance Quality • User’s Applications • Designing High Performance Systems and Scalable Architecture • Congestion Control Techniques New QoS Technologies • Migration to CISCO Technology • Interoperability – Middleware • Session Mobility & Integration 9
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
The four fundamental QoS viewpoints. Present & Future Value gap
Historic
Current Customer’s QoS Requirements
Alignment gap
Service Provider’s Offered QoS
Network Measurement & Monitoring IT Survey
Customer’s Perceived QoS
Customer
Perception gap
Execution gap
Service Provider’s Achieved QoS
Service Provider 10
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
QoS Definition • QoS is the ability of network element (e.g. an application, host or router) to have some level of assurance that its traffic and service requirements can be satisfied. • To enable QoS requires the cooperation of all network layers from to-to-bottom, as well as every network element from end-to-end. • Any QoS assurances are only good as the weakest link in the chain between sender and receiver. 11 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Quality of Service: QoS IT Survey
User Specified QoS
Application Specified QoS Media Quality - Media Relations
20 FRAME/sec - 640X320 PIXELS - 256 COLORS
Transmission and Operating Communications HP System Specified QoS OpenView Quantitative Criteria (throughput, delay, error-rate) Qualitative (synchronization, error recovery, scheduling, etc.)
Network QoS Parameters Requirements on Network Resources
Bandwidth Net. Load (nodal service units) Performance (jitter, cell-loss) Traffic Contract Parameters
Multimedia Devices QoS Timing and throughput demands
Frame-rate
delivery
12 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
User Perspective
The QoS Venn diagram
Application Perspective
Response Time
Image size, Color depth Voice quality, Steady picture, etc.
Transmission cost
Delay Jitter
Bandwidth Throughput, Burstiness, Compression, Transport technique
Delay, Jitter, Skew, Error rate
Throughput, Delay. Delay variance, Error rate
Transmission Perspective SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
13
The QoS performance dimensions required by some common applications Performance Dimensions Application
Sensitivity to
Bandwidth Delay
Jitter
Loss
VoIP
Low
High
High
Med
Video Conferencing
High
High
High
Med
Streaming Video on Demand
High
Med
Med
Med
Streaming Audio
Low
Med
Med
Med
E-Business (Web browsing)
Med
Med
Low
High
E-mail
Low
Low
Low
High
File Transfer
Med
Low
Low
High 14
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Application Features •
Collaboration across distributed heterogeneous platforms with:
Q o S
– Adaptability • Session mobility – switching between different terminal devices & display resolutions » black & white » color 640x480, 800x600, 1280x1024 or 1600x1200 pixels
– switching between different service providers & locations of servers » reduced latency
– Scalability • An application is scalable if its performance metrics can improve, as necessary and essentially without limit, by adding equipment (more hosts, more network connections, etc) and without the need for replacement of existing equipment. • Further, equipment cost should increase at most linearly with performance metrics, so that the cost per unit measure of performance is Messerschmitt 2000 constant or declines.
– Reliability • System ability to deliver continuous quality service 15 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
P r o v i s i o n
The Enhanced Disconfirmation Model of Customer Satisfaction
Quality = Customer’s Expectation – Customer’s Perception 16 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Determinants of Performance Quality Factor
Description: Completion time versus throughput
Software The architecture of the application determines the architecture partitioning of the application onto multiple hosts, which determines processing resources. It also addresses the relative location of data and processing. These both impact the communications overhead. Technology
Electronics, magnetic and optical storage, and fiber optics are rapidly advancing. Application performance benefits directly from these technology advances.
Equipment
Computing systems and network exploits technology to provide processing, storage, and communications services to applications. Their performance is determined by both technology and their internal architecture and design.
Messerschmitt 2000
17 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Case Study: Staffordshire University Network • Staffordshire University Campuses on large geographic area – Stoke-on-Trent; Stafford; Litchfield & Adjunct Colleges – Over 18000 university plus adjunct colleges’ users • Students; Academics; Technical & Administrative Staff
– Schools • Arts; Computing; Engineering, Business, etc.
• System Description & IT Architecture – Initial Design (5 years ago) • Simple Hub extended start topology based network • No QoS technology
• Data Collection – Network Measurements & Monitoring • MRTG data collection • HP Probe measurements
– IT Survey Formulation & Administration
• Data Analysis and Result 18 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Case Study:
Staffordshire Network Architecture
19 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
20 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Percentage
Utilization/Total Error 21 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
IT Survey study area of customer perception of QoS
22 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Score Off-campus
3.7
Student residence
4.2
On-campus
5.5
The results of the survey show that the majority of average scores on scale Excellent=10, Very Good=8, Good=6, Reasonable=4, Poor=2, Very poor=0, lie between 4 and 6, indicating that respondents typically see response time lying between reasonable and good, on average. The only exceptions appear to be the use of ftp transactions, which are often seen as less than reasonable, particularly at a student residence or at home. 23 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Correlation studies • There was a correlation between the user’s perception and the overall network utilization – With the increased level of utilization the user’s satisfaction is decreasing.
• The measurements suggest that the half-duplex network utilization was 70%; • The users’ perception was within the limits of reasonable and good; 24 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Conclusions I
• Discussion of : – – – – – –
Research Aims and Objectives QoS Fundamentals Users’ Applications Features Enhanced Disconfirmation Model Determinants of Performance Quality Case Study: Staffordshire University Network • Network Measurements and IT Survey • Discussion of Results 25
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Conclusions II
• Recommendation to IT Services: – The effective use of new QoS Technology • ATM, RSVP network architectures with capacity assignments based on application and transport level QoS parameters. • Implement more dynamic QoS provision technology – Session Mobility
– Apply more effective network monitoring policies • Proactive Measurements policies • Students Access Practices • Network Security, Convergence and Adaptability
– Maintaining users’ satisfaction • Competitive edge • Cost effectiveness 26 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
THANK YOU ANY QUESTIONS? 27 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Designing High Performance Systems and Scalable Architecture • • • •
•
•
Break the application down into small, atomic tasks, where each task will be assigned to one host (to avoid communication overhead. Characterize the resource requirements of each task. For example, how much processor execution time and memory and storage space does it consume? Analyse the communication pattern among tasks and in particular the communication burden on the network created by assigning tasks to different hosts. Understand the scheduling constrains on the task. What task depends on the prior completion of others (i.e., parallelism, scheduling?) In light of these constraints, tentatively assign tasks to different hosts in a way that attempts to achieve the maximum parallelism and minimizes communication requirements. Tasks prioritisation for what task should be completed at what host most urgently to reduce the communications overhead. Analyse the proposed host assignment in terms of hot spots to points of congestion in either processing or communication, as well as, under-utilised resources. What bottlenecks will ultimately limit overall application performance (i.e., response time to users)? 28
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Messerschmitt 2000
Congestion Control Implementation Techniques Approach
Description
Source Initiated
Sources detect network congestion, or are informed by the network (directly or through pricing), and limit the offered traffic. This has the advantage that sources can offer the most compelling traffic or shift less compelling traffic to another time.
Network Policing
The network enforces limits on the traffic it will accept from each source. The network may engage in flow control with each source, or it may silently drop excess traffic.
Messerschmitt 2000
29 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
IT Network Current Developments • New 34 Mbps Automatic-Fail-Over resilient link was added in parallel to the ATM 622 Mbps link between Stafford and Stoke campuses to deal with heavy traffic. • There is a proposed 34 Mbps circuit between Stoke and Lichfield site as well for the year 2003. • The E-mail server sits on very powerful Mellor 5509 switch in Stoke – Switch Octagon 5509 has CISCO Interconnection Operating System (IOS) version 12 installed with fast Ethernet port and Gigabit uplink services.
• The Web-server configuration on the Octagon 5509 switch and the web-cache is in Stoke • Use of the “H drive” to store the students’ scholarly files.
30 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
The IT Measurements Policies: • Reactive (i.e. ad-hoc proactive activities): Old policies – This is due to the lack of personnel training for using exploiting the technology to its maximum.
• Active: New Policies – The Packeteer implementation purely on RESNET only starting from September 2002, which is very active measurement of the type of traffic at student residencies • Due to high volume of complaints about slow ftp service due to the fact that more than 50 % of traffic was generated by the Audio-galaxy music files transfers. • The packeteer solution will help to identify and to shape the traffic accordingly in order to give higher priorities to the work related type of traffic (i.e., ftp, file sharing) instead of the entertainment type of traffic (music, games, etc.)
• New DHCP implementation requiring all the users to register their MAC addresses from the DHCP in order to be allowed to get the dynamic IP address required connect to the IT network. 31 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Discussion: Different QoS Technologies QoS Monitoring and Management
ApplicationSignaled QoS
RSVP
IP QoS
IP Differentiated Services (DiffServ)
NetworkSignaled QoS
ATM, PNNI, MPLS RSVP-TE or MPLS CR-LDP
Traffic Engineered Path
ATM PVCs, MPLS Lable Switched Path (LSPs)
Link Layer QoS
Ethernet 801p, VLAN, ATM, MPLS, PPP, UMTS, DOCSIS, Frame Relay
Physical QoS
Wavelengths, Virtual Circuits (VCs), Ports, Frequencies
Layer
32 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Discussion: IT Survey Results Campus Base Usage Frequencies
Respondent Type Frequency Percent Academic Staff
58
12.5
Clerical/Admin Staff
29
6.3
Management Staff
21
4.5
328
70.8
27
5.8
463
100.0
Student
Technical Staff Total
Frequency Percent Litchfield
31
6.7
Other
12
2.6
Stafford
277
59.8
Stoke-onTrent
143
30.9
Total
463
100.0
33 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Diagram representing the common usage of IT network
450 400 350 300 250 200 150 100 50 0
Usage Main usage
Univ campus
Student Res
Home 34
SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Remote Connections 160 140 120 100 80 60 40 20 0 128K Modem
56K Modem
ADSL
DSL
ISDN
35 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Type of Operating System
Number
Window s NT
Window s 2000
Window s 95
Linux 0
20
40
60
80
100
120
140
36 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Daily Usage Pattern 350 300 250 200 150 100 50 0 12-3am 3-6am
6-9am
912-3pm 3-6pm 12noon
6-9pm 9pm-12
37 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
User’s Session Duration Frequency More than two hours
Percent 199
44.1
Up to 10 minutes
12
2.7
Up to 30 minutes
38
8.4
Up to one hour
101
22.4
Up to two hours
101
22.4
Total
451
100.0
38 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
Respondent Application Usage admin app
email First use Usage web
ftp
0
20
40
60
80
Percentage
39 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng
User’s Perception of the Application Performance Campus
Student Residence
Home
Last use
5.1
4.4
4.4
ftp
4.0
2.8
3.3
Web
4.8
5.4
4.3
Email
5.5
5.4
4.6
Admin
5.0
4.1
40 SimpoTIC’03 BRATISLAVA OCTOBER 26-28, 2003
by Dr. Eur. Ing. Eduard Babulak PhD C.Eng