1. Network Design Principles
CP3397
Network Design and Security
Lecture 2

2. Contents Design goals Design choices Design approaches
The design process
Capacity planning

3. Design goals Good designs should: Deliver services requested by users
Deliver acceptable throughput and response times
Be within budget and maximise cost efficiencies
Be reliable
Be expandable without major redesign
Be manageable by maintenance and support staff
Be well documented

4. Design Choices Balance of distribution Level of transparency Security
Connectivity technology

5. Design approaches Two typical methods
Traditional analytic design
Building block approach
Both use a similar iterative approach

6. The traditional design process

7. Design Stages - Agree requirements
Engage end users
Translate requirements
Business objectives –> technical specification
Phasing the requirements
Right level of detail at each design stage
Designing the requirements

8. Design Stages - Designing the requirements
Aim for completeness
Prioritise with a hierarchical system such as
[M] - Mandatory
[H] – Highly desirable
[D] - Desirable
[N] - Note

9. Design Stages - Assessing requirements
Consider all aspects
E.g. support & maintenance, depreciation, commissioning costs, project management fees, h/w & s/w upgrade costs, b/w/ costs, consultancy charges – over the lifetime of the network
Weighted matrix multipliers
M=100, H=10, D=1, N=0
Produce scores and rank suppliers

10. Design Stages - Information gathering
Need to find details of user behaviour, application use and location information for example:
User: location, numbers, services used, typical access
Sites: number, location, constraints on traffic (security, political or cost)
Servers and services: location, level of distribution
WAN/backbone predicted link traffic
Protocol support: bridged, routed or switched – Gateways needed?
Legacy support: equipment, protocols or services
Specific availability needs? 24-hour/backup links etc
Five-year plan – changes to population or business requirements
Budgetary constraints
Greenfield or existing site
Information is refined and leads to a requirements database and capacity plan

11. Design Stages - Site constraints
Greenfield or
Greenfield sites have no legacy constraints but…
It is difficult to determine the real network loads and stresses
Needs more detail of application use and underlying protocols
Could use simulation to predict performance
Existing site
Limited access
Access to live network could be restricted but…
Bottlenecks more obvious
Can use traffic/network analysis tools

12. Design Stages - Planning
Uses information on
Hosts, users, services, and their internetworking needs
Iterative process of
Conceptual design
Analysis
Refinement
Involving
Brainstorming, design reviews, modelling tools
Leading to final draft design

13. Design Stages - Design specification
Detailed document of the design
Acts as a benchmark for design changes
Final design choices and changes need justification and documenting
Should include change history to aid maintenance
Used for the implementation

14. Design Stages - Implementation
Needs a project plan to include
Phased introduction of new technology
Educating the users (what to expect)
Pilot installation (test for possible problems)
Acceptance testing (to prove performance meets requirements)
Deployment (provide support on going live and provide fallback position)

15. Connectivity options Technology choices
LANs (Ethernet, Token ring, ATM)
MANs (FDDI, SMDS, ATM, SONET/SDH)
WANS (Frame relay, ATM, ISDN, X.25, PDCs, Satellite)
Wireless (802.11, Bluetooth, GPRS, GSM)
Dial-up lines
Serial links

16. Connectivity option determinants
Packet, cell or circuit switching
Wired or wireless
Distance
Performance
Bandwidth
Quality of Service
Availability

17. Media and bandwidth choices

18. Capacity Planning - Outline
Concerned with
User response times
Application behaviour and performance characteristics
Network utilisation
Needed to
Minimise downtime
Maximise service to customers
Minimise costs of procurement and maintenance
Avoid unscheduled maintenance or re-design
Avoid costly upgrades and bad publicity

19. Capacity Planning - Stages
Form a discussion group (involve users etc.)
Quantify user behaviour
Quantify Application behaviour
Baseline existing network
Traffic profiles
Make traffic projections
Summarize input data for design process
Assess other data (environmental, location restrictions, deployment constraints etc)

20. Capacity Planning – Step 1
Form a discussion group (involve users etc.)
Needs wide representation
Users, network managers, application groups
To elicit
What uses find acceptable and unacceptable
Map of services and users and details of user behaviour
Quantify items using
User and service sizing data
Snapshots from data capture and network management tools
Traces of key services using protocol analysers
Pilot network implementation

21. Capacity Planning – Step 2
Quantify user behaviour
Need to know population and and location of users
Summary of major user groups
Application use by user group
Site location data (country, grid ref., town, postcode, telephone exchange)
Planned changes

22. Capacity Planning – Step 3
Quantify Application behaviour
Need to identify
Applications that could affect performance
Location and performance of servers and clients
Key constraints on performance (response times, buffer sizes etc
And define
Application behaviour under fault conditions (lost data)
Addressing mechanisms( broad/multi/unicast)
Packet characteristics (frame sizes and direction)
Routable and non-routable services (IP, NETBIOS)
Undefined applications allow choice of distribution balance

23. Capacity Planning – Step 4
Baseline existing network
Baselining – a behavioural profile of the network obtained from
Packet traces, transaction rates, event logs and stats
Router ACLs, firewall rulebases
Inventory of H/W and S/W revisions
Traffic profiles -Capture data for a stable working network with details of
B/w utilization by packet type and protocol
Packet/frame size distribution
Background error rates
Collision rates
Various tools can be used
Network and protocol analysers, SNMP data, RMON probes, OS tools, traceroute, ping etc

24. Capacity Planning – Step 5
Make traffic projections using some, or all of:
Hand calculation
Commercial analytical tools to project network utilisation
Simulation tools (most detail)

25. Capacity Planning – Step 6
Summarize input data for design process
Budget
Database of sites, user populations,
List of key applications and their behaviour
Traffic matrix
Need to consider
Static or dynamic bandwidth allocation
Max. Delay and Max. hops between sites
Resilience, Availability, degree of meshing
Design constraints and trade-off
(e.g. delay v cost)

26. The building-block design process (an alternative)

27. Summary Good design Is an iterative process of continuous refinement
Is logical and consistent
Should deliver acceptable performance and cost metrics (trade-off)
Is more than choosing the technology!