section 5.2.

Data Center
Technology in the book Cloud

Reference book:Computing Concepts, Technology

& Architecture Thomas Erl, Zaigham Mahmood, and
Ricardo Puttini
Overview of Data Center Technology

• Definition: Data centers are centralized facilities that house computing resources, such as servers, storage,
and networking equipment, used to support cloud computing services.
• Key Components:
• Servers: Physical or virtual machines providing computing power.
• Storage Systems: Infrastructure for data management and storage.
• Networking: Equipment for communication within the data center and with external networks.
• Power and Cooling: Essential for maintaining operational stability.
Evolution of Data Centers
• Traditional Data Centers: Focus on physical hardware,
on-premises deployment, and manual management.
• Modern Data Centers: Shift towards virtualization,
automation, scalability, and cloud integration.
• Data center technology is crucial for cloud computing
but comes with significant challenges.
Challenges in Datacenters
1. Scalability Challenges
• Description: Scaling traditional data centers requires significant investment in hardware
and space.
• Implications: High costs and extended deployment times hinder responsiveness to
dynamic business needs.
2. Management Complexity
• Description: Managing a data center involves complex tasks like resource allocation,
maintenance, and security.
• Implications: Requires skilled personnel and sophisticated tools, increasing operational
complexity and costs.
3. Energy Consumption
• Description: Data centers consume a large amount of power, leading to high operational
costs and environmental impact.
• Implications: Growing focus on energy-efficient designs and the use of renewable energy
sources.
4. Security Concerns
• Description: Data centers must protect sensitive data from
breaches and unauthorized access.
• Implications: High stakes for compliance with regulations and
the implementation of robust security measures.
• 5. Data Center Downtime
• Description: Downtime can be caused by power failures,
hardware malfunctions, or cyberattacks.
• Implications: Leads to service disruption, financial loss, and
potential damage to reputation.
Emerging Solutions
• Automation and AI: Reducing manual intervention in
data center management.
• Software-Defined Data Centers (SDDC): Abstracting
hardware into software, improving flexibility.
• Green Data Centers: Focusing on sustainability and
energy efficiency.
Data Center Network Hardware
• Data centers require extensive network hardware for multi-level connectivity.
• Five main network subsystems form the core of data center networking
infrastructure.
1. Carrier and External Networks Interconnection: Provides routing between
external WAN connections and the data center's LAN.
2. Web-Tier Load Balancing and Acceleration: Web-Tier Load Balancing and
Acceleration
3. LAN Fabric: Facilitates network communication within the data center,
supporting speeds up to 10 Gbps and performing virtualization functions.
4. SAN Fabric: upports the implementation of storage area networks (SANs).
• Recall: SAN is is a high-speed, specialized network that provides block-level access to storage
devices, allowing servers to connect to storage arrays and manage data as if the storage devices
were locally attached.
5. NAS Gateways: Provides attachment points for NAS-based storage devices .

• Recall: Network Attached Storage (NAS) device is a file-level

computer data storage server connected to a computer network providing
data access to a heterogeneous group of clients.
 Bare Metal Servers
• A bare metal server, also known as dedicated servers, is a
physical computer server that is used by one consumer, or
tenant, only, without any virtualization.
• built to run 24/7 and usually have more reliable hardware. Some
have redundant power supplies, hard disks, etc.
• They come in different forms and shapes, the tower servers are
often used in office buildings
• When you buy a physical server, you have to think of the required resources. How
much memory does the server need? How much disk space? How fast should the
CPUs be? etc. You also have to take future growth into account
• Rack servers are placed in server cabinets:
• Later, blade servers were introduced that offer even more computing power / memory and use less
space
• The image below shows a blade enclosure that fits in a server rack. It offers networking, power, and
cooling. The blade servers have CPUs, memory, and storage. They fit in the slots of the blade
enclosure.
 Server Virtualization

• Nowadays, we use a lot of server virtualization which means

that we run multiple virtual machines on a single physical
server.
• All hardware of the VM (virtual machine) is virtualized, its
CPU(s), RAM, hard disks, network card, etc. For each virtual
machine you create, you can decide how many CPU(s), RAM,
etc. it will have.
• The virtualized server runs on hypervisor.
 Virtual Networking
• Physical servers have one or more NICs
(Network interface card) that are
connected to a network switch.
• Virtual machines have virtualized
hardware that includes their NICs, which
we call vNIC (Virtual NICs).
• these vNICs should be connect to our
network, this is done with a virtual
switch.
• Above you can see that the virtual
machines each have a vNIC that is
connected to a virtual switch. The virtual
switch is connected to a physical switch
through the physical NIC of the server
that the hypervisor runs on.
• In the picture above there is only one physical NIC. In
production networks, we usually use two or more
physical NICs for redundancy and to make sure there is
enough bandwidth for all virtual NICs.
• The virtual switch is supplied by the hypervisor vendor
or you can use an external virtual switch product like
the Cisco Nexus 1000v switch. This allows you to use
the same switch features on your virtual switch as you
use on your physical switches.
Physical Data Center Network
• We just talked about how virtual machines are connected to
a physical switch through a virtual switch.
• There are two different network switch deployment
strategies used in data centers. These strategies determine
the physical placement of network switches and how servers
connect to the network.
• the two approaches are :
• End of Row" (EoR)
• "Top of Rack" (ToR)
TOR (Top of Rack)
• The top of rack designs has network switches at the top of each server
cabinet / rack. The servers are located below the ToR switches and for
redundancy reasons, connected to both ToR switches
• The ToR switches are connected to distribution layer switches.
• One of the advantages of this setup is that most of your cabling remains
within the rack, the only network cables that leave the rack are those
from the ToR switches to the distribution layer switches.
• One of the disadvantages is that you need quite some ToR switches
and depending on how much servers you have in your rack, not all switch
ports on the ToR switches will be used.
EOR (End of Row)
• With the end of row design, there are no switches in the racks
and all servers are directly connected to EoR (End of Row)
switches that are located in a separate rack:
• Some of the advantages of this setup is that you don’t need as
many switches, there are less unused switch ports so overall port
utilization is better. One of the disadvantages is that you need a
lot of cabling from your server racks to the racks where the EoR
switches are located.
WAN Traffic Path to Cloud Services
• the different options we can use to connect to the cloud:
1. Internet:
• Using the Internet to connect to the cloud is a common option.
• we see that all virtual machines are located at the cloud provider.
• The enterprise network doesn’t have any servers or virtual machines
anymore, only users that require access to the applications that run on
the virtual machines.
• The advantages of using the Internet as your WAN connection to the cloud are:
• Cost: Internet access is cheap compared to private WAN options.
• Availability: it’s easy to get an Internet connection and it’s available almost everywhere.
• Migration: want to switch from one cloud provider to another? All cloud providers are
connected to the Internet so you don’t have to switch connections.
• Mobile users: if you have a lot of mobile users then they will be able to access your
applications whenever they have an Internet connection.

• Some of the disadvantages:

• Security: the Internet is a public network so it’s not a very safe place. Attackers might attempt
man-in-the-middle attacks to snoop on the traffic between your users and the cloud.
• Bandwidth: depending on the type of applications and the number of users, your Internet
connection might not have enough bandwidth for all users to access their applications.
• QoS (Quality of Service): the Internet is best-effort only, there is no quality of service. If you
have any applications that are sensitive to delay and/or packet loss then you might run into
issues.
• SLA: most Internet providers don’t offer any SLAs (Service Level Agreements) that guarantee a
certain bandwidth or availability. If you outsource all or most of your applications to the cloud,
you will be very dependent on your Internet connection.
• Most cloud providers also support IPsec VPN, allowing you to create a site-to-site VPN
tunnel between your Enterprise network and the Cloud provider.
2. Private WAN
• this is a dedicated connection from your site to the
cloud provider.
• Advantages:
• Bandwidth: private WAN connections offer a higher
bandwidth than most Internet connections.
• SLA: does offer service level agreements that guarantee a
certain bandwidth and availability.
• Disadvantages:
• Cost: private WAN connections cost more than regular
Internet connections.
• Availability: takes time to install the new connection.
• Flexibility: you are stuck to one cloud provider.
• some examples of private WAN connections:
• Microsoft Azure ExpressRoute
• Amazon AWS Direct Connect
3. Intercloud Exchange
• These are providers that are connected to multiple cloud providers and
offer you a private WAN.
• This solution is better if
• if you want to use multiple cloud providers and the advantages of private WAN
connections
• if you want to migrate from one cloud provider to another
• The intercloud exchange can offer you a connection to one or more
cloud providers without having to switch your private WAN connection
4. Virtual Network Functions
• Virtual Network Functions (VNFs) are a fundamental component
of Network Function Virtualization (NFV).
• They represent the various network functions that run on
virtualized infrastructure within a network such router, firewall ,
load balancer,switched, IDs,IPsor other network devices.
• VNFs enhance WAN traffic paths to cloud services by providing
optimized, secure, and flexible network functions.