AWS Series: #1 Cloud Deployment Models (public, private, poly & multi-cloud)- Part 2
This section of the article covers Hybrid Cloud, Poly Cloud, Multi Cloud, Shared Cloud Services & Modern Cloud Data Platforms. Refer to Part 1 of this series @ Cloud Deployment Models Part 1.
Hybrid Cloud:
What is it? Hybrid cloud platforms are a combination of both Private / On-premise and Public cloud. Not all organization’s regions / clients / country regulations allows to put all the workloads and data into cloud. Also not all the work loads can be migrated to public cloud directly in case of heavy lifting re-architecting and refactoring is needed. Such separation of workloads either before migrating to public cloud or due to any security / regulatory / latency restrictions is the birth of Hybrid cloud in any organization.
How does it work? Unlike Multi-cloud, Hybrid cloud is more homogeneous. usually a private cloud and a choice of public cloud. For any of these hybrid cloud, solutions such as AWS Outposts, Azure Stack, Google Anthos, VMWare’s VMConSAWS etc. works well for the hybrid cloud model.
What problem does it solve? Mainly Regulatory, Security and latency needs are the main reason for having a separate private cloud and having a public cloud.
Poly Cloud:
What is it? If we have the best of the clouds that has more than 100+ services provided by each of the major players, then what is the need to have Poly Cloud? Poly-Cloud is an architecture pattern to pick the best / dominant resources from each of the public cloud provider and build a multi cloud environment. The selection of the multi-cloud for this purpose is not random but dominant features or what best works for the firm in terms of team independence and providing efficient work environment. Usually Poly Cloud is accidental and is not a strategic approach. Good practice is for any firm who goes for single or multi-cloud approach, include poly-cloud as part of the architecture.
AWS: While there are 200+ products and services available as a subscription model with AWS, there are few flagship products that will make your architecture top-notch. Services such as most of the serverless products such as Lambda for processing / applications, Glue for ETL, DynamoDB for Databases and Redshift for Data Warehousing. Storage using S3 is a hit as that is the base for Data Lake and provides extensive storage possibilities.
Azure: Also with Azure there are several services available as subscription model. Microsoft Servers is the key flagship product. While Microsoft server is available with other cloud providers as well, provisioning it from Azure is better. Also Azure is enabling and investing more solutions for IoT providers.
GCP: GCP’s main focus is ML / AI and Containers. They have expanded the concept of containers especially Kubernetes for several years that makes it quite a dominant product for Google.
How does it work? Identifying differentiating features and and to pick what works for your organization and fit for the application or platform is the key. There is no one-size-fits all. If you are building a storage centric application ML specific application or high performance application that enables low latency or enabling multi-region access for distributed access depends on what is the center focus of your application, that is supposed to drive the application / platform which is the key for picking the dominant feature.
What problem does it solve? Architecture based on best or dominant solution each of the cloud providers offer and also enabling development teams to be able to independently and effectively work. Works best if Poly-Cloud is not an after thought.
Multi-Cloud:
What is it? Having heterogeneous cloud model (multiple private or public clouds) for an organization is a very normal thing, especially bigger organizations in order to provide more reliability, resilience, backup and availability, having a multi-cloud option is a norm of the day. Also, regulations of the client and several other regions based restrictions of the organizations or as a back-up strategy, organizations support multi-cloud. Also in order to avoid vendor lock-ins, use of dominant features from other cloud providers (PolyCloud), multi-region support with which ever regions has CDN is also a reason to opt Multi-Cloud solution.
How does it work? There are two options to create Multi cloud.
Option 1: Multi public cloud
Option 2: Multi Public + Private cloud
Refer to “Dedicated Private cloud” to read about AWS Outposts, Azure Stack and GCP Anthos.
What problem does it solve? Different flagship products from different public cloud providers, avoiding vendor-lock in, region based edge servers, client / organization regional restrictions to use specific cloud eg. use of Alibaba or GCP only,
Shared-cloud Services:
What is it? SAAS is mainly shared cloud services model. SAAS models such as Salesforce, Microsoft PowerBI, Tableau, Monday, Regulatory applications, Finance Data providers such as Bloomberg, Reuters etc. are shared cloud services which has Multi-tenant architecture.
How does it work? It is a multi-tenant architecture and internally either based on different databases / store or is separated by schema or low level separation or locks.
What problem does it solve? Instead of installing applications on-premise or building applications in-house subscribe to military-grade solutions available online saves time, effort and is easily managed.
Modern Cloud Data Platforms
What is it?
Snowflake: Integrated Data Warehousing and Data Science platform with the storage and compute separated. Storage can be either AWS or Azure and compute directly at the query level and the output can be stored back to Data Lake for further analysis and application of machine learning algorithms.
Refer: snowflake.com
Amazon: As this series focuses on AWS, separating the Data stack.
Each of the elements in the stack will be discussed in the future parts of AWS Series. Respective links will be amended in the future.
Azure: Azure Data Factory, Cosmos, SQL Server and data pipeline.
GCP: GCP Big Query & Data Lake.
Oracle Cloud: Oracle cloud, Exadata and Oracle Cloud Heatwave.
Databricks: Delta lake and opensource Lakehouse Architecture.
How does it work? Most of the Modern Cloud platforms not only provides PAAS and managed Databases but also provides integration to complete platform solution such as Analytics, Business Intelligence and Reporting. This enables the organization to build sophisticated Data Platforms with utmost security, reliability, resiliency and fault tolerance.
What problem does it solve? Managing Databases is a complex task. Maintenance and Administration with defined SLA and fixing the block corruption nightmares and taking backup / recovery is complicated and takes a lot of time. In order for organizations to focus on the business functionality and ability to build modern data platforms / products / services — Modern Data Cloud platforms is your friend.
Next article is about AWS cloud models management, AllOps (DataOps, MLOps, AIOps, BaseOps etc).
