AWS Series: #1 Cloud Deployment Models (public, private, poly & multi-cloud) — Part 1
In this article covers the cloud deployment models and which option may be more suitable for your organization. While IAAS, PAAS, SAAS and Serverless are Cloud models available on cloud provided by the solution providers, it depends on individual organization and the need to review and architect the deployment model.
Usually bigger organizations cannot just live with one model of public or private cloud, it mostly will be a combination of several cloud models and cloud providers in order to
(1) use best in class provided by each of these cloud providers — poly cloud
(2) in order to navigate the client / regional restrictions adopt private + public cloud model — hybrid cloud
(3) use of multiple cloud partners for production and backup differently or multiple cloud partners due to regional / client based restrictions — multi cloud
(4) use of different cloud models such as shared services (SAAS), PAAS, IAAS or serverless
(5) Organizations may also opt for dedicated private cloud provided by the public cloud providers so that the maintenance and administration tasks are handed over to the cloud providers and can have better focus on business needs. Security and Latency — Private / Dedicated private cloud.
(6) Moving from the legacy databases into the modern cloud databases. No big company shall use a single database model. It will always be a multi mix of Relational / Non-Relational / Graph as it will have volume and variety of data such as structure, semi-structured and unstructured. Also with the advent of Data Lake and Delta Lakes, building baby data lakes from the very beginning whatever the size of the organization is, makes it useful for organizations to grow and adopt. — Modern Cloud Data platforms.
Public Cloud:
What is it? It is like a Condo that provides space and services, ringfenced with tight security for condo and all additional facilities like Gym, swimming pool, tennis court, event rooms etc. each condo will provide differentiation feature based on who are its target audience and what kind of ecosystem it wants to offer. While umbrella Condo security service is provided each block and each unit is further ringfenced with security and at some level it is a shared responsibility.
Similarly, a Public cloud provides computing and storage service at the bare minimum with additional services such as security, infrastructure, networking, sourcing, streaming, servers, databases, file storage, containers, load balance, machine learning, big data and analytics, monitoring, messaging, backup & recovery and more. Depending on the cloud provider, they offer both server-based and serverless offerings. It is up to the organization whether they want to put all the eggs in the same basket or want to distribute it across different cloud providers. Also, there will be flagship services offered by different cloud providers. Upon using such services makes the final product much more efficient and effective.
Of the multiple cloud providers, Amazon Web Services (AWS), Microsoft Azure and Google Cloud Provider (GCP) are the main public clouds. There are other cloud providers such as Oracle Cloud, Snowflakes, IBM, Alibaba etc. which provides exemplary services. Also, the army of cloud services and expertise must be part of the organization’s cloud stack to be readily used as the need to use it may occur any time including but not limited to compliance and regulatory reasons of the client / organizations regional presence.
How does it work? Public cloud provides storage, compute and other services and it works based on the subscription model.
What problem does it solve? For bigger organizations there is not much difference between On-Premise and cloud. Starting from Security to covering all the OSI layer, similar things needs to be done on cloud as well. However, there are couple of key differentiating factor between On-Premise and Cloud. For smaller and start-up organizations, cloud is a boon where they can have a quick head-start and can be quick to market with their MVP with fraction of investments that they would make for on-premise set up.
For Infrastructure:
(1) Ownership model to subscription model (2) Hosted in sharable Data Centre and be aware of infra of public cloud.
For Application Development:
(1) Security-First for Application Development (2) Shared architecture and possible noisy neighbors (3) Best-in class cloud-only resources (4) Use of more managed services which will give more time to deliver functional requirements.
For Business:
(1) Possibly quick to market as servers can be procured sooner and horizontal scalability is possible.
For Finance / Operations:
(1) CAPEX to OPEX for for the firm. Certain architecture could still be CAPEX. (2) Depreciation cycle maintenance and approach differs.
Private Cloud:
What is it? Having own private Data Centre and creating a public cloud like architecture including scalability, resilience and fault tolerance with auto healing and continuous availability of the servers that are equivalent to the promise delivered by public cloud. For much sensitive organizations where they do not want to put all / part of their data and resources on public cloud, Private cloud gives them the provision to park it yet have all the facilities of the private cloud. Organizations also use private cloud as a Refactoring opportunity and use it as a staging environment for 2–3 years before they migrate the work load to the public cloud. How does that impact the below groups?
For Infrastructure team:
Weight is not lifted off the Infrastructure team by enabling Private cloud. Infact from On-Premise to moving to the Private cloud model there are so many advantages and increases the maintenance routine for the Infrastructure as additional layer of containerization and auto recovery, auto healing, ensuring reliability, resiliency and 100% availability must be ensured. There is a overlap in the work between On-premise to Private cloud and some more.
For Application Development team:
For the Application Development team, since the infrastructure and provisioning is enabled by infra team, their work is more or less the same when compared to on-premise. However, if Private cloud is adopted for the purpose of getting out of legacy applications and to use modern application stack, there is a lot of work in re-architecting the solution, refactoring etc., based on the migration methodology that is to be adopted. The projects will take several years to complete hence parallel run in the on-premise environment will be required and the cost will be doubled during this period. Especially from On-premise to Private cloud model, the servers will have to be either bought / leased hence there will be additional heavy cost for bigger organizations if they choose this model. But many organizations prefer the Private Cloud approach as they are able to effectively rebase their strategy and provides bandwidth and runway for efficient refactoring of applications.
How does it work? To build private cloud, the organization can use its own servers or public cloud providers to build dedicated private cloud. Also Redhat’s Openstack, helps to build and manage pool of virtual environments and provides the reliability, availability and resilience. It is a virtualization management platform. There are other providers such as HPE private cloud enables the management of Private cloud for the organizations.
What problem does it solve? Many organizations have sensitive data that they do not want to migrate to public cloud. There could also be regional restrictions for organizations where private is preferred over public or there could be client related restrictions where they want their data to be represented in a private environment. All the problems are sorted by having private cloud for all or partially. Private cloud are also increasingly seen as an essential first step before migrating to public cloud as it gives more migration time and options to refactor / repurchase / retire systems. When there is a strong need for security, regulations and latency needs — Private cloud is the way to go.
Dedicated Private Cloud:
What is it? Private cloud could be set up by self or public-cloud providers could provision their own servers and resources to set up dedicated private cloud for the firm. It involves network, switching, servers, complete infrastructure, complete security protocol etc., to be provisioned for the organization by the public cloud providers. This is also pretty costly when compared to Public cloud. However, the servers can be managed by the public cloud providers which may provide more bandwidth to focus on new business requirements and refactoring legacy applications into better modern platforms. All that was discussed for private cloud applies to dedicated private cloud.
How does it work? Major provides solutions to build dedicated private solutions such as AWS Outposts, Azure Stack, Google Anthos, VMWare’s VMConSAWS etc. These public cloud providers enables you to build private cloud based on the servers and services provided by them.
AWS Outposts: You will be able to run AWS products and services on premise using AWS Outposts. Fully managed service that provides the same infra, service, tools, API’s etc. as you use it in the AWS public cloud. Provides a consistent, reliable and resilient hybrid approach.
Azure stack: Like AWS Outposts, Azure stack enables organizations to access the Microsoft Azure products and services from on-premise.
Refer: AWS Outposts vs Azure Stack.
GCP Anthos: Anthos is a managed platform that extends the Google cloud services to your own on-premise environments.
Refer: AWS Outposts vs Azure Stack vs Google Anthos
Azure Arc: In recent times Microsoft is focusing on the Multi-cloud and integrations. Azure Arc has been there around for a while. Azure Arc provides simplified management and easy development environment for multi-cloud. Provides easy ability to AWS to Azure to GCP.
What problem does it solve? It solves all the problems a Private cloud could solve. Additionally, it also solves the problem of management and administration of servers, networks, switches and have shared-security etc.
Part 2 of the story covers Cloud Deployment Model — Part2.
