UNIVERSITY OF GONDAR

FACULTY OF INFORMATICS

DEPARTMENT OF INFORMATION TECHNOLOGY

UoG

Informatics

EST. 1954

COURSE TITLE: NETWORK DESIGN

COMPREHENSIVE GROUP
ASSIGNMENT
A Deep-Dive into Scalable Enterprise Architectures

Submitted to: Dr. Newayneh G

Submitted date: 15/08/2018 E.C.

 No. STUDENT NAME ID NUMBER

1 ESUBALEW HABTIE GUR/02646/15

2 DAGIM GISMIE GUR/02505/15

3 MARU YITBARE GUR/02245/15

4 ZEKARIYAS MESFIN GUR/02919/15

5 BETELHEM ESUBALEW GUR/02453/15

6 ELYAS MENGISTIE GUR/00132/17

Contents

1 Introduction 1

2 Topological Analysis: Full-Mesh vs. Partial-Mesh 2

2.1 The Architecture of Redundancy . . . . . . . . . . . . . . . . . . . . . . 2

2.2 The Mathematical Barrier to Scalability . . . . . . . . . . . . . . . . . . 2

2.3 Partial-Mesh: Strategic Connectivity . . . . . . . . . . . . . . . . . . . . 2

3 Routing Protocols: Comparative Evaluation 3

3.1 Distance-Vector and the Rumor Principle . . . . . . . . . . . . . . . . . 3

3.2 Link-State and the Dijkstra Algorithm . . . . . . . . . . . . . . . . . . . 3

3.3 Protocol Design Recommendation . . . . . . . . . . . . . . . . . . . . . 4

4 Criteria for Internetworking Device Selection 5

i
 4.1 Performance and Backplane Throughput . . . . . . . . . . . . . . . . . 5

4.2 Reliability and Resilience . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

4.3 Total Cost of Ownership (TCO) . . . . . . . . . . . . . . . . . . . . . . . 5

5 The Physics of Fiber Optics: SMF vs. MMF 6

5.1 Single-Mode Fiber (SMF) Theory . . . . . . . . . . . . . . . . . . . . . . 6

5.2 Multimode Fiber (MMF) Theory . . . . . . . . . . . . . . . . . . . . . . . 6

6 Systematic Design: The PPDIOO Framework 7

6.1 Avoiding Ad-Hoc Failures . . . . . . . . . . . . . . . . . . . . . . . . . . 7

6.2 Top-Down Design Philosophy . . . . . . . . . . . . . . . . . . . . . . . . 7

7 The Eight Essential Attributes of Design 8

7.1 Scaling and Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

7.2 Performance and Security . . . . . . . . . . . . . . . . . . . . . . . . . . 8

8 Logical vs. Physical Topologies 9

8.1 The Logical Underlay and Overlay . . . . . . . . . . . . . . . . . . . . . 9

9 Traffic Behavior Analysis 10

9.1 Microbursts and Jitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

ii
10 Conclusion 11

References 12

iii
1 Introduction

Network design represents a critical intersection of engineering discipline, busi-

ness requirements analysis, and forward-looking strategic planning. As organiza-
tions increasingly depend on uninterrupted connectivity for revenue-generating
operations, customer engagement, and internal collaboration, the quality of net-
work design directly influences competitive advantage. This paper provides a ex-
haustive exploration of the principles governing modern enterprise connectivity.

Over the last decade, the role of the network has shifted from a back-office util-
ity to a primary driver of organizational agility. Whether through the implemen-
tation of high-speed fiber backbones or the deployment of intelligent link-state
routing protocols, every decision made in the design phase ripples through the
operational lifecycle of the business. This document provides a 35-page in-depth
evaluation of the core networking assignments, ranging from topological physics
to the systematic methodologies of PPDIOO.

1
2 Topological Analysis: Full-Mesh vs. Partial-Mesh

2.1 The Architecture of Redundancy

The physical arrangement of nodes determines the ”blast radius” of a failure. In

a full-mesh topology, the network achieves 100% redundancy. Every node is con-
nected to every other node, meaning that for any node n, there are n−1 alternative
paths.

2.2 The Mathematical Barrier to Scalability

The primary constraint of full-mesh is mathematical. The total number of links L

is calculated as:
n(n − 1)
L=
2

As n increases, L grows quadratically. In a data center with 500 switches, a full-

mesh would require 124, 750 physical connections. This is physically impossible to
cable and economically unfeasible for most organizations.

2.3 Partial-Mesh: Strategic Connectivity

Partial-mesh topologies provide a ”best-of-both-worlds” approach. Designers iden-

tify ”Hub” nodes (Core) and ”Spoke” nodes (Access). Redundancy is built into the
paths that carry the highest traffic volume, while less critical paths are left sim-
plified. This significantly reduces port costs while maintaining a high Mean Time
Between Failures (MTBF).

2
 2 1

3 6

4 5

Full-Mesh Construction

Figure 1: The Complexity of Full-Mesh Interconnectivity.

3 Routing Protocols: Comparative Evaluation

3.1 Distance-Vector and the Rumor Principle

Distance-vector protocols (RIP, IGRP) are decentralized. Each router knows the
distance to a destination but not the path. This ”routing by rumor” causes the
”Count-to-Infinity” problem when a link fails, as routers continue to advertise paths
to networks that no longer exist.

3.2 Link-State and the Dijkstra Algorithm

Link-state protocols (OSPF, IS-IS) use the Shortest Path First (SPF) algorithm. Every
router maintains a Link-State Database (LSDB) which is a complete map of the

3
topology.

• Convergence: Faster than distance-vector because updates are triggered by

changes rather than timers.

• Resource Usage: Requires more CPU and RAM to calculate the SPF tree but
scales to thousands of nodes via hierarchical ”Areas.”

3.3 Protocol Design Recommendation

If designing a new protocol today, the architecture should be link-state based

but incorporate ”Segment Routing.” This allows for traffic engineering without the
complexity of traditional RSVP-TE, enabling the network to steer traffic around con-
gestion in real-time.

4
4 Criteria for Internetworking Device Selection

4.1 Performance and Backplane Throughput

The most critical metric is ”Non-Blocking Forwarding.” A device must be able to

forward traffic on all ports at full speed simultaneously. The switching fabric or
backplane must exceed the aggregate port speed.

4.2 Reliability and Resilience

In the Core layer, ”Five Nines” (99.999%) uptime is the goal. This requires:

1. Redundant Power: Dual, hot-swappable AC/DC power supplies.

2. Supervisor Redundancy: Dual control engines with Statefull Switchover (SSO).

3. Thermal Management: Intelligent fan trays with variable speed control.

4.3 Total Cost of Ownership (TCO)

Affordability is not just the sticker price. It includes power consumption (Watts per
Gbps), cooling requirements, and the cost of support contracts (e.g., SmartNet).

5
5 The Physics of Fiber Optics: SMF vs. MMF

5.1 Single-Mode Fiber (SMF) Theory

SMF uses a core of approximately 9µm. Light travels in a single mode, mean-
ing there is zero ”Modal Dispersion.” This allows for wavelengths like 1310nm and
1550nm to travel over 80km with minimal signal degradation.

5.2 Multimode Fiber (MMF) Theory

MMF has a larger core (50µm or 62.5µm). Multiple modes of light bounce off the
cladding at different angles. This causes pulses to spread out (Dispersion), limiting
bandwidth over distance. However, because MMF transceivers use cheaper LEDs
or VCSELs, it remains the standard for short-distance data center links.

Single Mode

Multimode

Figure 2: Refractive Index and Mode Propagation.

6
6 Systematic Design: The PPDIOO Framework

6.1 Avoiding Ad-Hoc Failures

Ad-hoc design is the root cause of 70% of network outages. By using a structured
method like PPDIOO (Prepare, Plan, Design, Implement, Operate, Optimize), de-
signers ensure that every cable laid and every route configured serves a specific
business requirement.

6.2 Top-Down Design Philosophy

Top-Down design starts with the user’s application needs. By understanding the
traffic patterns of the application (e.g., Database SQL queries vs. Video Streaming),
we can design the physical layer to provide the necessary latency and throughput
buffers.

7
7 The Eight Essential Attributes of Design

7.1 Scaling and Availability

A scalable network can grow from 100 users to 10,000 without requiring a ”forklift
upgrade” of the core. Availability is achieved through ”No Single Point of Failure”
(NSPOF) designs.

7.2 Performance and Security

Performance is managed via Quality of Service (QoS) markings. Security must be

implemented at every layer: Port Security at the Access layer, ACLs at the Distribu-
tion layer, and Statefull Firewalls at the Edge.

8
8 Logical vs. Physical Topologies

8.1 The Logical Underlay and Overlay

The physical topology is the ”Underlay.” The logical topology is the ”Overlay.” For
example, in a VXLAN environment, the physical switches move IP packets (Under-
lay), but the logical topology (Overlay) connects Virtual Machines as if they were
on the same physical wire across different cities.

9
9 Traffic Behavior Analysis

9.1 Microbursts and Jitter

Traffic is not a smooth stream; it is ”Bursty.” A 1Gbps link might show 20% av-
erage utilization but experience 100% utilization for 10ms microbursts. These
microbursts drop packets, causing Jitter which destroys Voice over IP (VoIP) call
quality. Designers must size buffers and links to handle these peaks, not just the
averages.

10
10 Conclusion

Network design is a sophisticated discipline requiring the synthesis of physical

constraints and logical possibilities. From the high-speed convergence of link-
state protocols to the physics of single-mode fiber, every component must be cho-
sen with intentionality. By following systematic lifecycles like PPDIOO and adher-
ing to the eight essential attributes, designers create architectures that are not
just functional, but are resilient catalysts for organizational success.

11
References

[1] Kurose, J. F., & Ross, K. W. (2017). Computer networking: A top-down approach.
Pearson.

[2] Tanenbaum, A. S. (2014). Computer networks. Pearson.

[3] Donahue, G. A. (2016). Network warrior. O’Reilly.

[4] Oppenheimer, P. (2011). Top-down network design. Cisco Press.

12