Understanding Private Cloud on the Cloud platforms

A lot of choices when it comes to choosing the cloud providers

Hello everyone, hope all of you are doing well and keeping healthy and safe. I am kicking off these series of blogs on understanding cloud offerings by different cloud providers. There are at least three ulterior motives behind this and I am hoping it becomes apparent as you read through these articles.

Ever wondered why Google, Amazon and IBM calls something VPC (virtual private cloud) on GCP, AWS and IBM cloud respectively, while Oracle calls it VCN (virtual cloud network) on OCI and Microsoft calls it VNET (virtual network) on Azure. 

Well, not all of us may know it as they might practically be the same feature in the cloud to provide an isolated network in the cloud in a shared network infrastructure but there might still be subtle differences between how these private cloud environments are orchestrated by these different cloud providers. This is the first and foremost objective of this blog and that is to break the myth around the private cloud. In order to accomplish this, we will proceed in a step by step manner –

Understanding network IP and subnet addressing

Internet Protocol Addressing

An IP address serves two main functions. It identifies the host and location of the host in the network and thereby an ability to establish a path to that host.

• An IPv4 address has a size of 32 bits and that is because and that is because the IPv4 address in dotted-decimal notation has 4 – 8 bits separated by dots.
• These 4*8 comes to 32 bits which limits the address space to 2³² IPv4 address space which is approximately ~ 4.29 billion addresses.
• Out of these 4.29B addresses, some are reserved: all 0s and 1s, private address, broadcast and so forth. Remaining Addresses are known as Public IP Addresses (~4B).
• Since each IP address has 4*8 bit combination, each 8 bit can expand to 2⁸ bit IP beginning with zero to 255.

Public IPv4 addresses are globally unique and can be reachable from anywhere. That’s why these global addresses have a DNS configured that interprets the domain name to a public IPv4 address. Routing works by delivering packets to a destination network and the organization that owns the destination network is responsible for internally routing/delivering the packet to a machine.

Three classes of network were defined as per internet protocol RFC791:

Classless Inter domain range (CIDR)

• Classless inter domain routing (CIDR) was introduced to accommodate a flexible addressing scheme to conserve address space.
• Number of bits used to identify network is explicitly stated with   /<number> notation
• Address is allocated based on organization’s actual need
• IPv6 also uses CIDR Notation and more comprehensively handles address space shortage
• Example of CIDR notation –

Subnets

• Network are more or less sub-divided into subnets inside an organization
• This subnetting of address assists in manageability of network, security, isolation etc.
• CIDR block convention is followed to identify subnets

A private cloud configuration on a cloud platform requires you to specify a CIDR and a subnet CIDR.

Private Address Space

• Private address is a reserved space (RFC1918)
• Organizations are free to use this space for their own internal private network
• These addresses cannot be used for public addresses

You can read more about IP addresses and usage from the WIKI page. I have added my interpretation here for clarity and also to build a background for what we want to discuss about virtual private clouds.

Identifying Overlapping CIDR ranges

You may always run into a situation where you are configuring VPCs and need to accurately identify if the CIDRs overlap before you can peer the VPCs or use some other networking options to peer them. I am have enlisted a few examples below in order to demonstrate a simplification of identifying overlapping CIDRs.

I think you all may already know this but let me document it so that we all understand what I have understood how to identify starting and ending addresses quickly in CIDR range. The subnet mask in a CIDR range in a way identifies what parts of the 4 – 8 bits will change in CIDR.

Take this example – 192.168.0.0/16

The most significant 16 bits – highlighted in RED – identifies the network – 11000000.10101000.00000000.00000000

The remaining 16 bits identify the variable portion of the network for hosts and since each 8 bit is 2⁸ number of hosts, the ending address of CIDR can span to 255 max on last 16 bits.

Now, that we have few basic networking concepts out of the way, lets look at the private cloud and networking options of different cloud providers. I have tabulated the data below and provided links to the specific cloud providers documentation for every feature identified below.

I am sincerely hoping that this documentation to serve a few specific purposes –

1. This should help the readers identify the commonalities of design and networking concepts behind creating a private cloud on a cloud providers platform.
2. We still need a collective voice from the certification community as well as customers deploying these multi-cloud patterns to ensure that these cloud providers work towards common nomenclature of service offerings on their platforms. This will definitely help adoptions faster. It has to be understood that the more exclusivity in these naming conventions – just to sound and look different – will not only increase the time in demystifying a service offering but also increase the chances of falling behind a competition as well as a larger probability of extinction.
3. If you are preparing for certification for a specific cloud provider platform, you might as well use this documentation to prepare for certification for others cloud providers. This will become more and more relevant as you see organizations choosing multiple cloud providers instead of just selecting one and get locked.

There is also another motive behind this documentation. Please see our effort to create a global data center map of the five cloud providers.

As organizations mature towards multiple cloud platforms it become necessary that these cloud providers have some clairvoyance towards a globally unified world where these data centers are created with more globally distributed locations versus being concentrated in specific geographic locations. This will connect and unify the world especially in times of dire need like what we are going through!!

That’s all I have for today. I plan to bring similar perspective to other cloud options like storage, databases (relational or otherwise), big data analytics, security options, IaaS, PaaS or SaaS offerings. I hope all of you find this useful.

References/Links:

• GCP Locations
• Azure Cloud Locations
• Oracle Cloud Locations
• AWS Cloud Global Infrastructure
• IBM Cloud Data Centers
• Managed Sentinel – Documentation of On-Premises versus Cloud

If you are interested in other cloud certifications and how to prepare for them, check out our website.

Have questions? Participate in Cloud discussions on our Forums. Click here.

More from same author:

• Cloud 101 for mainframe developer
• Practical Comparisons of DynamoDB and VSAM
• How to Pass OCI Associate Architect Exam
• How to Pass OCI Professional Architect Exam

AUTHOR: Mukesh Sharma is a multi and hybrid cloud enthusiast with a bias for building robust hybrid cloud systems around mainframes for financial services organizations. You can reach him on LinkedIn