Logical Network Design and Technology Choice Notes

Background and Context of Logical Network Design

• Logical design is a critical phase in the Network Analysis and Design (NAD) process, occurring after Requirement Analysis and Flow Analysis.

• It builds directly upon flow information gathered during previous stages, including:

  • Requirement Specifications: Encompassing application, user, and host requirements.

  • Flowspec: Containing capacity plans and service plans, which detail the performance characteristics of individual, composite, and backbone flows.

• The logical design phase serves two primary functions:

  • Making technology selections for the network.

  • Developing an interconnectivity plan.

• This phase also facilitates the development of network-layer routing and addressing strategies, which subsequently incorporate location information to drive the Physical Network Design (e.g., cable plant design and equipment placement).

Process Model for Logical Design

The process follows a specific 6-step sequence:

1. Step 1: Establish Design Goals: Aligning technical objectives with organizational needs.

2. Step 2: Developing Criteria for Technology Evaluation: Setting the rules for how candidate technologies will be measured.

3. Step 3: Making Technology Choices: Selecting specific technologies based on capacity and service plans (from the Flowspec).

4. Step 4: Apply Interconnection Mechanisms: Determining how different parts of the network will interface.

5. Step 5: Integrated Network Management and Security: Planning for visibility and protection.

6. Step 6: Risk Analysis and Contingency Planning: Preparing for potential failures or limitations.

Step 1: Establishing Design Goals

• Precise specification of design goals is essential; without them, there is no objective way to distinguish success from failure except by user opinion.

• MINIMAX Principle: A universal set of design goals that reflects applying "Value" to design.

  • Minimize Cost: Reducing deployment and operations expenses.

  • Maximize Performance: Optimizing speed, delay, and reliability.

  • Value Equation: $\text{Value} = \frac{\text{Performance}}{\text{Cost}}$

• Diversified Goals: Requirements may vary across the network; for instance, flow requirements change based on whether they are best-effort, specified, or guaranteed services. One size does not fit all.

• Common Design Goals:

  • Minimizing deployment costs (including initial hardware and circuits) and operations costs.

  • Maximizing specific network performance characteristics.

  • Ease of use and manageability.

  • Adaptability: Ensuring the network can adjust to new or changing user needs.

  • Optimizing Security: This can mean maximizing absolute security, mapping security to specific group requirements, or providing multiple security models within the same network.

• Design Space: Goals can act as thresholds (constraints) that cannot be crossed. The design space may be one-dimensional (e.g., speed) or multi-dimensional.

Factors Influencing the Design Environment

• Designing a Brand New Network: Frees the designer from existing constraints. Key factors include:

  • Scalable infrastructure.

  • High-performance backbones.

  • Connectivity for specialized workgroups.

• Designing for an Existing Network: Focuses on managing the transition. Key factors include:

  • Minimizing disruption to existing services.

  • Working within the constraints of existing infrastructure, protocols, and technologies.

Step 2: Developing Criteria for Technology Evaluation

Criteria for evaluating technology are translated from design goals and the Flowspec.

• Analytical Inputs:

  • Capacity planning and service planning from the Flowspec.

  • Environment-specific design goals (e.g., minimizing deployment costs, ease of use).

• Common User Criteria:

  • Use of Standards-based technologies.

  • Preference for Generally Available technologies.

  • Selection of Proven and Tested technologies, leading to COTS (Commercial Off The Shelf) products.

Dimensions of the Design Space

• Maximizing Performance: Selecting technologies that meet or exceed specific requirements for capacity, delay, and reliability. This often increases complexity and requires specified services in the design.

• Maximizing Reliability: Involves introducing redundancy through technology choice, network topology, and routing protocols.

• Adaptability: Awareness of dynamic system environments.

  • Virtual network concepts.

  • Rapid reconfiguration (user groups, address assignment, locations).

  • Dynamic Flow Management: Managing flows individually, often through switching technology.

Criteria 1: NBMA and Broadcast Technologies

Evaluations must distinguish between how different technologies handle traffic distribution:

• Non-Broadcast Multiple Access (NBMA):

  • Do not have an inherent broadcast mechanism.

  • Typically switched-circuit technologies such as ATM, Frame Relay, SMDS, and HiPPI.

  • Consideration must be given to how IP address resolution is supported (e.g., hardwiring address bindings or utilizing proprietary broadcast support).

• Broadcast Technologies:

  • Traditional LAN technologies like Ethernet, Token Ring, and FDDI.

  • Use standard IP address resolution protocols such as ARP and RARP.

• Trade-offs:

  • Simplicity: Broadcast is effective for local, directly connected access.

  • Scalability: Some NBMA technologies can handle broadcast flexibly by creating logical sub-networks independent of physical topology, which increases complexity and the need for sophisticated management.

Hierarchy as a Fundamental Principle

• Rationale: As groups grow larger, broadcast traffic becomes burdensome, eventually disrupting performance. Hierarchy handles this complexity.

• Hierarchy Levels: Can be organized by users, computers, applications, areas, or Autonomous Systems (ASs).

• Group Sizing Rule: Ensure that steady-state background broadcast traffic remains a small percentage, specifically < 2%, of the total capacity of the technology.

  • Example: In a $10\,Mb/s$ Ethernet network, the broadcast traffic threshold would be $200\,Kb/s$.

• Measurement: Group size is determined by the number of physical connections or the number of ports on a switch.

Connection Support and Network State

Connection support involves mapping addresses across the network (end-to-end like IP, or local like ATM virtual circuits) and managing state information.

• Hard State: Connection information is determined and maintained throughout the long-lived path between source and destination.

• Soft State: System knowledge is short-lived; it is maintained only until the connection is established or for a brief period thereafter.

• Stateless (No State): The network does not determine or maintain state information between source and destination (Connectionless technology).

• Trade-offs for Connection-Oriented (Hard/Soft State):

  • Cost: Higher overhead in terms of CPU, memory, bandwidth, and monitoring.

  • Control: Allows for more configuration control but results in delays due to setup and teardown times, especially for short-lived traffic.

Criteria 2: Functions and Features of Technology

• Adaptability: Technologies should help the system adapt to mobility, resource availability, or lack of existing communication infrastructure.

• Supporting Specified Requirements: Addressing asymmetric flows (where users typically receive more bandwidth) through mechanisms like:

  • Frame Relay CIRs (Committed Information Rates).

  • SMDS access classes.

  • ATM QoS (Quality of Service).

  • IP TOS (Type of Service) levels.

Capacity Upgrade Paths

• Technology selection must account for future scalability.

• SONET (Synchronous Optical Network) hierarchy is based on the base rate OC-1 (Optical Carrier-1).

  • OC-1 Calculation: $(9\text{ rows}) \times (90\text{ columns}) \times (1\text{ byte/row-column}) \times (8000\text{ frames/second}) \times (8\text{ bits/byte}) = 51.84\,Mb/s$.

• Available Upgrade Paths:

  • SONET (Concatenated - OC-Nc):

  • OC-3c: $155.52\,Mb/s$

  • OC-12c: $622\,Mb/s$

  • OC-48c: $2.488\,Gb/s$

  • OC-192c: $9.953\,Gb/s$

  • Ethernet: $10\,Mb/s, 100\,Mb/s, 1\,Gb/s, 10\,Gb/s, 40\,Gb/s, 100\,Gb/s$.

  • FDDI: $100\,Mb/s$ to $1\,Gb/s$.

  • HiPPI: $800\,Mb/s \rightarrow 1.6\,Gb/s \rightarrow 6.4\,Gb/s$ (Super HiPPI).

  • ATM: $1.5\,Mb/s$ to $622\,Mb/s$.

  • Frame Relay: $56\,Kb/s$ to $445\,Mb/s$.

Backbone Flows and Evaluation Guidelines

• Backbone Flows: These are principal points where flows are consolidated and hierarchies are established.

• Scaling Factor: Used to handle large hierarchy factors. A common scaling factor used in examples is 1.5.

  • Example: If backbone capacity is $3.15\,Mb/s$, and the scaling factor is 1.5, the Modified Backbone Capacity is $(1.5) \times (3.15\,Mb/s) = 4.725\,Mb/s$.

• Technology Evaluation Rules:

  • Rule 1: If specific services are specified in the Flowspec, the selected technology must support those specified services.

  • Rule 2: If the requirement is "best effort," the technology choice should be based upon capacity planning for each individual flow.

Example Design Environment Case Study

• Scenario: Local workgroup network.

• Budget: $\$2M$ for purchase of PCs and network (excluding installation/maintenance).

• Hardware Cost: PCs are $\$2500$ each.

• Quantity: Between $300$ and $600$ PCs depending on remaining network budget.

• Workgroup Size: Solutions computed in groups of $32$ to $64$ PCs.

• Performance Requirements:

  • Capacity: $80\,Mb/s$.

  • Delay: $100\,\mu s$ between any two PCs in a workgroup.

Questions & Discussion

• Q: How does hierarchy affect backbone flows?

• A: As the hierarchy factor increases (e.g., from 2:1 to 5:1), the capacity needs and performance probability requirements (e.g., 99.5% to 99.95%) also increase to maintain network stability.

• Q: What is the relationship between the Flowspec and Step 2?

• A: The Flowspec's capacity planning determines baseline capacity requirements, while service planning evaluates technology's ability to provide specified services like QoS or security.