Components of Cloud Computing

Overview of Cloud Infrastructure

After choosing the cloud service model and the cloud type offered by vendors, customers need to plan the infrastructure architecture. The infrastructure layer is the foundation of the cloud.

Region

It is a geographic area or location where a cloud provider’s infrastructure is clustered, and may have names like NA South or US East.

Availability Zones

• Multiple Availability Zones (AZ)
• Have their own power, cooling, networking resources
• Isolation of zones improves the cloud’s fault tolerance, decrease latency, and more

• very high bandwidth connectivity with other AZs, Data Centers and the internet

  

Computing Resources

Cloud providers offer several compute options:

• Virtual Servers (VMs)
• Bare Metal Servers
• Serverless (Abstraction)

Storage

Virtual servers come with their default local storage, but the stored documents are lost as we destroy the servers. Other more persistent options are:

• Traditional Data Centers:
  • Block Storage
  • File Storage
• Often struggle with scale, performance and distributed characteristics of cloud.
• The most common mode of storage
  • Object Storage
• It is highly distributed and resilient

Networking

Networking infrastructure in a cloud datacenter include traditional networking hardware like:

• routers

• switches

  For users of the Cloud, the Cloud providers have Software Defined Networking (SDN), which allows for easier networking:

• provisioning
• configuration

• management

  

  Networking interfaces in the cloud need:

• IP address

• Subnets

  It is even more important to configure which network traffic and users can access your resources:

• Security Groups
• ACLs
• VLANs
• VPCs

• VPNs

  Some traditional hardware appliances:

• firewalls
• load balancers
• gateways

• traffic analyzers

  Another networking capability provided by the Cloud Providers is:

• CDNs

Types of Virtual Machines

Shared or Public Cloud VMs

Transient or Spot VMs

• The Cloud provider can choose to de-provision them at any time and reclaim the resources

  These VMs are great for:

• Non-production
• Testing and developing applications
• Running stateless workloads, testing scalability

• Running big data and HPC workloads at a low cost

  

Reserved virtual server instances

• Reserve capacity and guarantee resources for future deployments

• If you exceed your reserved capacity, complement it with hourly or monthly VMs Note: Not all predefined VMs families or configuration may be available as reserved.

  

Dedicated Hosts

• Single tenant isolation
• Specify the data center and pod
• This allows for maximum control over workload placement
• Used for meeting compliance and regulatory requirements or licensing terms

Bare Metal Servers

A bare metal server is a single-tenant, dedicated physical server. In other words, it’s dedicated to a single customer.

• Cloud Provider manages the server up to the OS.
• The Customer is responsible for administering and managing everything else on the server.

Bare Metal Server Configuration

• Preconfigured by the cloud provider
• Custom-configured as per customer specifications
  • Processors
  • RAM
  • Hard drives
  • Specialized components

  • The OS

    Add GPUs:

• Accelerating scientific computation
• Data analytics
• Rendering professional grade virtualized graphics

Characteristics

• Can take longer to provision
• Minutes to hours
• More expensive than VMs
• Only offered by some cloud providers

Workloads

• Fully customizable/ demanding environments
• Dedicated or long-term usage
• High Performance Computing

• Highly secure / isolated environments

  

Bare-metal server vs. Virtual Servers

Bare Metal	Virtual Servers
Work best for: CPU and I/O intensive workloads	Rapidly provisioned
Excel with the highest performance and security
Satisfy strict compliance requirements	Provide an elastic and scalable environment
Offer complete flexibility, control, and transparency
Come with added management and operational over head	Low cost to use

Secure Networking in Cloud

Networking in Cloud vs. On Premise

To create a network in cloud:

• Define the size of the Network using IP address range, e.g.,: 10.10.0.0/16

  

Direct Connectivity

Building a Cloud

It entails creating a set of logical constructs that deliver networking functionality akin to data center networks for securing environments and ensuring high performing business applications.

Containers

Containers are an executable unit of software in which application code is packaged, along with its libraries and dependencies, in common ways so that it can be run anywhere—desktops, traditional IT, or the cloud. Containers are lighter weight and consume fewer resources than Virtual Machines.

• Containers streamline development and deployment of cloud native applications
• Fast
• Portable
• Secure

Cloud Storage and Content Delivery Networks

Basics of Storage on the Cloud

• Direct Attached/Local Storage
  • Within the same server or rack
  • Fast
  • Use for OS
  • Not suitable
    • Ephemeral (Temporary)
    • Not shared
    • Non-resilient
• File Storage
  • Disadvantages
    • Slower
  • Advantages
    • Low cost

    • Attach to multiple servers

      

• Block Storage
  • Advantages

    • Faster read/write speeds

      

• Object Storage
• Disadvantages
  • Slowest speed
• Advantages
  • Least expensive
  • Infinite in size

  • Pay for what you use

    

File Storage

Like Direct attached:

• Attached to a compute node to store data

  Unlike Direct attached:

• Less expensive
• More resilient to failure
• Less disk management and maintenance for user

• Provision much larger amounts of Storage

  File storage is mounted from remote storage appliances:

• Resilient to failure
• Offer Encryption

• Managed by service provider

  File storage is mounted on compute nodes via Ethernet networks:

Multiple Compute Nodes

• File storage can be mounted onto more than one compute node
• Common Workloads:
  • Departmental file share
  • ‘Landing zone’ for incoming files

  • Repository of files

    i.e., speed variance is not an issue

• Low cost database storage

IOPS

Input/Output Operations Per Second – the speed at which disks can write and read data.

• Higher IOPS value = faster speed of underlying disk
• Higher IOPS = higher costs
• Low IOPS value can become a bottleneck

Block Storage

What is Block Storage?

Block storage breaks files into chunks (or block) of data.

• Stores each block separately under a unique address.

• Must be attached to a compute node before it can be utilized.

  Advantages:

• Mounted from remote storage appliances
• Extremely resilient to failure

• Data is more secure

  Mounted as a volume to compute nodes using a dedicated network of optical fibers:

• Signals move at the speed of light
• Higher price-point
• Perfect for workloads that need low-latency
• Consistent high speed
• Databases and mail servers
• Not suitable for shared storage between multiple servers

IOPS

For block storage, as it is for file storage, you need to take the IOPS capacity of the storage into account:

• Specify IOPS characteristics
• Adjust the IOPS as needed
• Depending on requirements and usage behavior

Common Attributes of File and Block Storage

• Block and File Storage is taken from appliances which are maintained by the service provider
• Both are highly available and resilient
• Often include data encryption at rest and in transit

Differences: File Storage vs. Block Storage

File Storage	Block Storage
Attached via Ethernet network	Attached via high-speed fiber network
Speeds vary, based on load	Only attach to one node at a time
Can attach to multiple computer nodes at once
Good for file share where:
1) Fast connectivity isn’t required	Good for applications that need:
2) Cost is a factor	1) Consistent fast access to disk

Remember: Consider workload IOPS requirements for both storage types.

Object Storage

1. Object storage can be used without connecting to a particular compute node to use it:

1. Object storage is less expensive than other cloud storage options
2. The most important thing to note about Object Storage is that it’s effectively infinite - With Object Storage, you just consume the storage you need and pay per gigabyte cost for what you use.

When to use Object Storage:

• Good for large amounts of unstructured data
• Data is not stored in any kind of hierarchical folder or directory structure

Object Storage Buckets

Managed by Service Provider

Object Storage – Resilience Options

Object Storage – Use Cases

• Any Data which is static and where fast read and write speeds are not necessary
  • Text files
  • Audio files
  • Video files
  • IoT Data
  • VM images
  • Backup files

  • Data Archives

    Not suitable for operating systems, databases, changing content.

Object Storage – Tiers and APIs

Object Storage Tiers

Standard Tier

• Store objects that are frequently accessed
• Highest per gigabyte cost

Vault/Archive Tier

• Store objects that are accessed once or twice a month
• Low storage cost

Cold Vault Tier

• Store data that is typically accessed once or twice a year
• Costs just a fraction of a US cent per/GB/month

Automatic archiving rules

• Automatic archiving rules for your data
• Automatically be moved to a cheaper storage tier if object isn’t accessed for long

Object Storage – Speed

• Doesn’t come with IOPS options
• Slower than file or block storage
• Data in ‘cold vault’ buckets, can take hours for retrieval
• Object storage not suitable for fast access to files.

Object Storage – Costs

• Object Storage is priced per/GB

• Other costs related to retrieval of the data e.g., Higher access costs for cold vault tiers

  Ensure data is stored in correct tier based on frequency of access.

Application Programming Interface, or API

Object Storage – Backup solutions

• Effective solution for Backup and Disaster Recovery
• Replacement for offsite backups
• Many backup solutions come with built-in options for Object Storage on Cloud
• More efficient than tape backups for geographic redundancy

CDN – Content Delivery Network

• Accelerates content delivery to users of the websites, by caching the content in data centers near their locations.
• Makes websites faster.
• Reduction in load on servers
• Increase uptime
• Security through obscurity