The cloud infrastructure consists of data centers and the hardware that runs in them, including compute, storage, and networking hardware; virtualization software; and a management layer
The Physical Environment of the Cloud Infrastructure
Just like traditional or onsite computing, cloud computing runs on real hardware that runs in real buildings. At the contemporary scale of operations, data center design, cloud infrastructure and operation are unlike anything else.
The following are the characteristics the best cloud infrastructure
- There is a high volume of expensive hardware—up to hundreds of thousands of servers in a single facility.
- Per square meter, power densities can be found up to 10kW (kilowatts).
- There is an enormous and immediate impact of downtime on all dependent businesses.
- Datacenter owners can provide multiple levels of service. The basic level is often summarized as “power, pipe, and ping.”
- There is electrical power and cooling pipe that is air conditioning. “Power” and “pipe” limit the density with which servers can be stacked in the data center.
- Power density is expressed in kW per rack, where a data center can house up to 25 racks per 100 square meters.
- Power densities of 100W per rack were once the norm, but these days 10kW or more per rack is seen and often required to ensure adequate supply can satisfy operational and functional requirements. These densities require advanced cooling engineering
- Network connectivity is provided to the data center networks to access storage, and external connectivity is provided to access wide area network (WAN) resources.
- Datacenter providers (colocation) can provide floor space, rack space, and cages (lockable floor space) on any level of aggregation. The smallest unit can range from a 1U slot in a rack to a full room.
Given the low tolerance for failure, the physical environment of the data center should be evaluated for geographic and political risks (seismic activity, floods, availability of power, and accessibility).
Cloud Data Center Infrastructure
A large part of data center design and cloud infrastructure revolves around the redundancy in the design.
Anything that can break down should be replicated.
No single point of failure should remain.
This means backup power, multiple independent cooling units, multiple power lines to individual racks and servers, multiple power distribution units (PDUs), multiple entrances to the building, multiple external entry points for power and network, and so on
Network and Communications in the Cloud
The purpose of the network is to provide for and control communication between computers—that is, servers and clients. According to the National Institute of Standards and Technology’s (NIST’s) “Cloud Computing Synopsis and Recommendations,
” the following first-level terms are important to define:1
- Cloud service consumer: Person or organization that maintains a business relationship with and uses service from the cloud service providers (CSPs)
- CSP: Person, organization, or entity responsible for making a service available to service consumers
- Cloud carrier: The intermediary that provides connectivity and transport of cloud services between the CSPs and the cloud service consumers
In the NIST cloud computing reference model, the network and communication function is provided as part of the cloud carrier role. In practice, this is an Internet protocol (IP) service, increasingly delivered through both IPv4 and IPv6.
This IP network may or may not be part of the public Internet
Cloud Network Functionality and Infrastructure
Functionality in the network includes the following:
- Address allocation: The ability to be able to provide one or more IP addresses to a cloud resource via either a static or a dynamic assignment.
- Access control: The mechanisms used to grant or deny access to a resource.
- Bandwidth allocation: A specified amount of bandwidth provided for system access or use.
- Rate limiting: The ability to control the amount of traffic sent or received. Can be used to control the number of application programming interface (API) requests made within a specified period.
- Filtering: The ability to selectively allow or deny content or access to resources.
- Routing: The ability to direct the flow of traffic between endpoints based on selecting the best path.
Software-Defined Networking (SDN)
Software-defined networking’s (SDN’s) objective is to provide a clearly defined and separate network control plane to manage network traffic that is separated from the forwarding plane.
This approach allows for network control to become directly programmable and distinct from forwarding, allowing for dynamic adjustment of traffic flows to address changing patterns of consumption.
SDN enables you to execute the control plane software on general-purpose hardware, allowing for the decoupling from specific network hardware configurations and allowing for the use of commodity servers.
Further, the use of software-based controllers permits a view of the network that presents a logical switch to the applications running above, allowing for access via APIs that can be used to configure, manage, and secure network resources.
For example, an SDN service might allow Internet access to a certain server with a single command, which the SDN layer can map to configuration changes on multiple intermediate network components.