INTRODUCTION

💻 1️⃣ The Internet & Data Centers — What’s Going On

Over the past few decades, the Internet has turned from a small research network into a massive
global system that runs everything — from your Google search to your Instagram feed.
Behind all that, the real work happens inside data centers — huge buildings full of servers that
process and store information.

🏢 2️⃣ Types of Data Centers

There are two main types:

1. Service-Oriented Data Centers — give online services directly to users.

🔹 Examples: Google, Facebook, Yahoo.
→ You log in → they provide search, social, mail, etc.

2. Resource-Oriented Data Centers — give computing power and storage to other users or
companies.
🔹 Examples: Amazon EC2, Microsoft Azure.
→ You rent servers or storage instead of buying your own.

In short:

Google = uses its own data for users.

AWS/Azure = rents data center power to others.

⚙️3️⃣ Why Data Centers Matter

 Companies no longer keep hundreds of small local servers.

 Instead, they consolidate everything into centralized hubs (data centers).

 These massive centers are the backbone of cloud computing — every cloud service (Gmail,
Netflix, ChatGPT, etc.) runs inside one.

So as cloud computing grows, data centers become essential — and optimizing their performance is
a key challenge.

4️⃣ The Purpose of a Data Center

A good data center must:

 Process huge amounts of data quickly, and

 Manage the entire IT infrastructure centrally (servers, storage, network, power, cooling).

But this isn’t easy — building and maintaining them is complex and expensive.

📈 5️⃣ Growth of Data Centers

  Around 1994, data centers started expanding rapidly.

 Today, they are classified into four tiers (I–IV) — based on how reliable and fault-tolerant
they are.

🧱 6️⃣ Tier Classification Explained (Table 10.1)

Uptime
Tier Description
(%)

Basic setup, no redundancy. A single failure can bring it down. Used by small
Tier I 99.671
businesses.

Tier II Tier I + some backup systems. Can handle small issues. 99.741

Tier
Tier II + multiple power/network connections. Can run even during maintenance. 99.982
III

Tier Fully fault-tolerant — everything (power, cooling, storage, network) is duplicated.

99.995
IV Used for mission-critical servers like banks, hospitals, cloud giants.

Tier IV = maximum reliability

Tier I = simplest and cheapest

So when you hear “Tier IV Data Center,” think maximum uptime, no single point of failure.

🌐 7️⃣ Why Small Companies Don’t Build Data Centers

Maintaining one is too costly and too technical.

So smaller companies rent resources from the big players (Google, Amazon, Microsoft, etc.).

That’s literally what cloud computing is — using someone else’s data center instead of building your
own.

⚡ 8️⃣ What Big Companies Use Data Centers For

Companies like Google, Amazon, Facebook, etc., use their data centers for:

 Web search

 Social media

 Web hosting

 E-commerce

 Cloud storage

 Big data computations

Basically, every online service you use daily runs on a data center.
⚠️10️⃣ Current Challenges

Even with all the tech advances, data centers still face big challenges:

 High server and infrastructure costs

 Excessive power consumption

 Network bottlenecks (slow data flow inside centers)

So the focus now is on:

 Building faster, more efficient networks, and

 Optimizing the transport layer (how data moves between servers).

🔹 10.2 Overview of Data Center Environment – Simplified Notes

🏢 Early Stage – Server Rooms
 In the beginning, organizations used server rooms to store their servers
and networking devices for LAN (Local Area Network).
 Some companies maintained:
o One main server room for normal operations, and

o One standby server room (backup) in case the main one failed.

 To ensure fault tolerance, a few even placed these rooms in different

buildings to avoid total shutdown during disasters.
In short: early data centers were just basic, isolated server rooms with minimal
fault tolerance.

🌐 Rise of Data Centers

 With the growth of client–server computing and the Internet, the concept
of data centers was born.
 Data centers became essential for providing high reliability and continuous
availability of online services.

⚙️Role and Evolution

 Data centers became the core foundation of IT systems — supporting and
even becoming the business itself.
(For example, Google or Facebook — their entire business is their data center.)
  The design of data centers has been evolving for decades, but in the past
5–10 years, the pace of innovation has accelerated sharply due to
advances in cloud computing, virtualization, and automation.

🧱 Mission-Critical Nature
 The term “mission critical” means that the hardware and software in data
centers must never go down — uptime is vital.
 A data center acts like a fortress:
o Highly secure

o Designed for maximum reliability

o Prioritizes zero downtime, even at high cost.

In short: reliability is the top priority, but modern data centers are also
becoming more efficient, automated, and innovative.

Perfect — this section is all about understanding how classical (traditional)

data centers are designed.
Here’s a clean, exam-focused explanation of 10.2.1 Architecture of Classical
Data Centers — simplified and structured for memory retention 👇

🔹 10.2.1 Architecture of Classical Data Centers – Simplified Notes

🏢 Definition and Purpose

 A data center is the central hub where all computing, storage, and
applications of an organization operate.
 It supports enterprise-level IT systems — like ERP (Enterprise Resource
Planning), CRM (Customer Relationship Management), and e-commerce.
 Every piece of information either comes from or passes through the
data center.
Key design goals:
✅ Performance – fast and efficient processing.
✅ Resiliency – must remain operational even during failures.
✅ Scalability – must support business growth and higher load.
⚙️Types of Classical Data Center Architectures
There are two major models used in enterprise data centers:
1️⃣Multitier Model (Most Common)
 Based on three logical layers (tiers):
1. Web Tier – handles client (browser) requests.
 2. Application Tier – processes business logic.
3. Database Tier – stores and retrieves data.
Used in:
 Enterprise systems like ERP and CRM (e.g., Oracle, Siebel).
 Web-based applications built with Microsoft .NET, Java 2EE, etc.
🔐 Advantages
 Resiliency: If one server fails, others in the same tier continue the work.
 Security: Attackers breaching the web server cannot directly reach the
database.
 Scalability: Each tier can be scaled independently (add more web
servers, etc.).
⚙️How it Works
 Each tier can run on:
o The same physical machine (connected using IPC – Interprocess
Communication), or
o Different machines (connected via network).

 Typically, large enterprises use separate servers for each tier (called
server farms) to handle heavy traffic efficiently.
Example:
You open an online shopping website →
Web server shows the product page →
App server handles your order logic →
Database stores your order details.
2️⃣Server Cluster Model
 Originated from universities and research labs for scientific
computation, but now used in industries like:
o Finance

o Manufacturing

o Entertainment

Associated with:
 HPC (High-Performance Computing) – large-scale parallel processing.
 HTC (High-Throughput Computing) – handling large volumes of smaller
tasks.
 Grid / Utility computing – distributed resources shared over networks.
Key idea:
A group of servers (a cluster) works together as one powerful unit, often running
customized or proprietary applications for specific business goals.
⚡ Comparison: Multitier vs Cluster Model

Aspect Multitier Model Server Cluster Model

Web and enterprise apps (ERP,

Purpose Scientific, financial, HPC tasks
CRM)

Structur
Web → App → Database tiers Group of servers working in parallel
e

Scalabili
Scales per tier Scales horizontally by adding nodes
ty

Application performance &

Focus Raw computational power
security

Weather simulation, animation

Example Banking website
rendering

🔹 10.2.2 Cloud-Enabled Data Centers (CEDCs) – Simplified Notes

⚙️Background / Need
 Modern businesses face technological and economic pressure — they
need more performance from their existing IT infrastructure.
 Traditional data centers are struggling to handle:
o Increased number of users

o Larger applications

o Higher performance demands

 This leads to inefficiency and reduced service quality.

☁️Solution: Cloud Computing
 Cloud computing offers a solution by:
o Reducing costs (no need to buy or maintain heavy hardware)

o Increasing flexibility (resources can scale up or down as needed)

o Allowing on-demand resource management

🧠 What is a CEDC?
CEDC = Cloud-Enabled Data Center
It is an evolved form of a traditional, virtualized data center — made
smarter and more automated.
It combines virtualization, automation, and cloud service layers (IaaS,
PaaS) to make the data center intelligent and dynamic.

🔄 How It Works
 Takes the concept of virtualization (where hardware resources are
divided into virtual machines)
 Adds intelligence + automation using IaaS and PaaS frameworks
⚡ Key Features of CEDC
1. Agility: Quickly adapts to workload or business changes.
2. Automation: Automatically allocates and manages computing resources.
3. Integration: Manages different (heterogeneous) data centers as one
unified system.
4. Optimization: Provides exactly the resources required — improving
efficiency.
5. Flexibility: Supports both physical and virtual systems seamlessly.
🎯 Business Benefits
 Better efficiency and cost control
 Improved performance and resource utilization
 Simplified management across multiple data centers
 Faster response to changing business needs

10.2.3 Physical Organization 👇

Data centers can occupy an entire building, a few floors, or even just one big room. Inside, there are
many servers arranged neatly in rack cabinets. These racks are placed in rows, leaving aisles
(corridors) between them so people can access the front and back of each rack easily.

Some large storage devices are as big as the racks themselves and are placed next to them.
Big data centers can have thousands of servers, and sometimes they use shipping containers filled
with about 1000 servers each. When something fails or needs an upgrade, they replace the whole
container instead of fixing individual servers.

10.2.4 Storage and Networking Infrastructure – Short Explanation

Data centers need several types of networks because different devices communicate in different
ways.
There are mainly four types of physical networks used inside a data center:

1. Client–Server Network:
Connects the data center to users on the Internet.
Uses wired Ethernet or Wi-Fi for external access.
 2. Server–Server Network:
Allows high-speed communication between servers inside the data center.
Uses Ethernet or faster systems like InfiniBand.

3. Server–Storage Network:
Connects servers to storage devices where data is kept.
Often uses Fibre Channel, or sometimes Ethernet or InfiniBand.

4. Management Network:
Used to monitor and manage the data center.
Usually Ethernet, but with separate cabling for safety.

Inside a data center, multiple applications run on different (often virtual) servers.
Each application has its own public IP address, and incoming user requests from the Internet are
distributed across servers by a load balancer — this keeps performance stable.

Most modern data centers use a two-tier network topology.

Here, access switches (that connect servers) are combined with aggregation switches (that handle
routing and switching).
This design has advantages like:

 Simple layout

 Faster communication (less latency)

 Fewer switches → less power use

But it also has one drawback: limited scalability.

When the aggregation switches reach full capacity, adding new ones becomes complex and creates
routing challenges.

10.2.5 Cooling Infrastructure – Short Explanation

Data centers generate a huge amount of heat because thousands of servers run continuously.
To prevent overheating, a cooling system is used that includes air-conditioning units, fans, and air
circulation ducts.

Servers are placed on raised floors (plenum) in alternating cold and hot aisles:

 Cold air is pushed upward in the front aisles.

 Server fans pull the cold air through and push hot air out the back.

 The hot air rises and is sent to the chiller plant for cooling and reuse.

This setup is efficient and not very expensive, but sometimes causes hot spots where cooling is
uneven.

Recent years have brought new energy-efficient cooling designs, using green technology and
creative layouts — like mobile container data centers, and buildings designed to act as natural air
handlers.
Major companies like Google and Microsoft use hyperscale facilities optimized for both power and
cooling.
Locations for data centers are chosen based on fiber network access, low disaster risk, and cool
climates for natural temperature control.

10.2.6 Nature of Traffic in Data Centers

The way data moves inside a data center (called traffic) is different from how it moves on the
Internet.
Data centers use private, high-speed networks with very low delay — often less than 250
microseconds.

Traffic inside data centers is divided into three types:

1. Mice Traffic:
Small requests like Google searches or Facebook updates.
Needs very quick responses and low delay.

2. Cat Traffic:
Medium-sized transfers like downloading small files or uploading photos.
Needs smooth and consistent speed.

3. Elephant Traffic:
Large transfers like software updates or video downloads.
Needs high bandwidth and stable throughput.

All these types of traffic happen together, so the network must handle small, medium, and large
data flows efficiently without slowing down.

10.3 Networking Issues in Data Centers – Simple Explanation

Data centers are the backbone of cloud computing — all cloud services depend on fast and reliable
networking.
However, maintaining smooth and secure communication between thousands of servers is not easy.
There are several major networking issues that cloud providers face.

1️⃣ Availability

Availability means that cloud services should be accessible all the time without interruption.
One of the biggest challenges for cloud providers is to maintain maximum uptime (near 100%).
Even a few seconds of downtime can cause:

 Loss of customer trust and reputation

 Violation of Service Level Agreements (SLAs)

 Financial loss to the company

To ensure high availability, cloud providers use:

 Data replication (keeping multiple copies of data in different locations)

 Regular backups (so data can be restored quickly if something fails)

Example: If one server fails, another copy of the same data on a different server automatically takes
over.

2️⃣ Poor Network Performance

Data centers handle different types of traffic — small (mice), medium (cat), and large (elephant).
Each type needs different speed and response levels, so the network must be flexible and fast.

The three key performance goals of a Data Center Network (DCN) are:

 High burst tolerance – ability to handle sudden spikes in traffic

 Low latency – very little delay in sending and receiving data

 High throughput – ability to transfer large amounts of data quickly

The problem is that traditional TCP/IP (the main Internet protocol) was designed for normal Internet
communication, not for the high-speed, complex environment of data centers.
As a result, it often causes:

 Delays (latency issues)

 Uneven traffic handling

 Poor performance under heavy load

Hence, cloud networks need improved or modified communication protocols to achieve the best
performance.

3️⃣ Security

Security is another major concern in data center networking.

The cloud provider must keep user data safe during:

 Transit (when data is moving through the network)

 Rest (when data is stored in servers)

Common security issues include:

 Data loss due to power failure or software malfunction

 Unauthorized access or insider threats

 Physical attacks on the data center or network equipment

To prevent these, providers use:

 Encryption (to protect data during transfer and storage)

 Regular backups and monitoring

 Physical security measures like surveillance, restricted access, and biometric locks