Multi cloud and edge computing are hotly discussed topics these days. To understand why that is, it is important to understand application infrastructure evolution and the impact of cloud on enterprises.

Application evolution

Applications have evolved from monolithic to virtualized multi-tiered to modern microservices based architecture. Application infrastructure has evolved from client-server to on-prem data centers to cloud based models, in parallel.

Newer use cases in many industry verticals are requiring applications to be geographically distributed for better user experience, cost, security and compliance. Examples include cloud gaming, remote asset monitoring, connected retail, digital healthcare and autonomous vehicles.

Such applications need distributed real-time decision making closer to end users and data sources, with feedback mechanism within the application span in order to be adaptive. This is now possible with the availability of affordable specialized compute (GPUs, TPUs) at the edges coupled with increased network speeds (5G). Such applications will have loosely coupled component functions that are geodistributed but are tightly connected.

Application architecture and infrastructure evolution

Enterprise digital transformation and Cloud

Over the last decade, enterprises have been going through digital transformation and Cloud has been an important catalyst for this. Cloud has revolutionized IT and made many enterprises agile in their application delivery. The cloud providers have created a modern architecture to consume IT as a service with SLAs that are otherwise hard to replicate.

It’s now very easy for a small company or a team in a large enterprise to swipe a credit card and get IT and application infrastructure set up. However, the cloud based IT consumption has also come at a cost for enterprises. This is perhaps more evident now as one can assess real impact over time.

• Economics - The article True Cost of Cloud provides some real data on impact on cloud on enterprises. It is estimated a staggering $500B+ market value has been lost by global enterprises due to cloud impact on margins. There have been examples of Dropbox increasing margins with infrastructure optimization, which is also similar to what was reported by Zoom in their most recent earnings report
• Skillset - When enterprises managed their own IT, the IT admins had different skillsets ranging from infrastructure to applications. This has changed to DevOps in the recent years, with “infrastructure as code” now becoming prevalent. In the process, most enterprises have lost the traditional IT skillset as it is now consumed as a service from the cloud
• Lock-in - Initially the cloud adoption starts with consuming compute and storage instances as IaaS, but companies constantly make short-term tradeoffs at every step and consume DBaaS, PaaS etc. due to the ease of use. Before they realize, they become more dependent on a particular provider. It’s almost like a quicksand. Cloud lock-in is the modern version of vendor lock-in by individual infrastructure and application component vendors a generation ago.

In addition, dominance of traditional IT vendors has significantly reduced in the last decade in favor of cloud providers who offer everything-as-a-service and have a faster pace of innovation.

Cloud v/s Datacenter spend by enterprises

Hybrid cloud, multi cloud and edge

So, economics is a big driver for enterprises to consider hybrid and multi cloud architectures going forward. In addition, data locality, compliance (GDPR) and proximity (latency), or just to reduce reliance on a single provider are important drivers for enterprises for developing distributed application architectures that are independent of providers.

What are the options for enterprises?

• Hybrid Cloud - repatriation from cloud to a private datacenter could be an option. However, it requires breadth of expertise to manage the datacenter, in addition to applications, which most enterprises don’t have today. So even for large enterprises, a partial repatriation could be more realistic.
• Multi Cloud - choice of cloud providers exist and this market will eventually become an oligopoly of small number of providers instead of a monopoly. While not all clouds are made equal w.r.t capabilities and user experience, top 3-5 cloud providers now have services that meet the needs for a large percentage of enterprises. 
• Edge - the emergence of micro datacenters by datacenter providers such as Equinix makes it possible to extend application closer to end users. This option could be applicable to both small and large enterprises

No matter the choice, distributing applications across clouds, datacenters and edges can be hard. The complexity and the unknowns often becomes a deterrent for most organizations.

To avail these options, it is important for the application infrastructure to be portable across providers and locations.

Application portability

Applications have evolved from legacy monolithic and 3-tier architectures to modern microservices (cloud-native) architecture. Depending on the age of the enterprise and where they are in their digital transformation journey, they tend to be on a spectrum of a mix of legacy and modern application infrastructure. 

The applications are made of components that perform specific functions (i.e. services or microservices) and designed with a framework to run at scale. The framework includes:

• Orchestration - To associate and deploy application components on compute instances for execution, with the ability to scale when needed
• Connectivity - For the components of an application to discover and interact with each other
• Data and Storage - To manage the data used and produced by the application
• Security - The security framework encompasses network, data, access and application level parameters
• Observability - A framework to monitor the application through events and alerts, and a mechanism for feedback and fixing issues
• Workflows and Lifecycle Management - Applications have workflows for performing the business logic and have a CI/CD framework for lifecycle management

The application portability depends on how this framework has been designed and what parts of the framework are portable.

In addition, the ease of portability also depends on whether the application components are designed as VMs or containers and if they are tethered to any local resources or services such as DBaaS etc. which make them sticky with a provider or location.

A practical approach

Application workloads can be portable more easily if they are designed as containers or serverless functions. Whereas, data portability needs more thought as it might have compliance, locality, network costs related considerations. This would require workload placement, connectivity and security framework design that is more agile and portable.

A roadmap needs to be defined to not only modernize the applications (if they are not already), but also application redesign for portability. The portability is often an evolution and not a 0 to 1 transition for most applications and enterprises. The benefits of application portability will outweigh the costs for enterprises.

And the benefits of leveraging hybrid or multi cloud or edge will far outweigh the costs in the medium to long run.