Tango in the clouds

We are at an inflection point (again)

Right shift -> Client-Server to Cloud Computing

After several years of churning out client-server architected products, the development of cloud architected products over the last couple of years was refreshing to say the least. The thinking required dismantling the existing client-server model and latching onto all things distributed. It does require a bit of effort, but once developed, the distributed paradigm with loosely coupled microservices gives you a completely new way to look at existing as well as new problems. The advent of API’s which make the cloud resources available for your applications to exist upon has changed the consumption economics forever.

There was a time when every household, town, farm or village had its own water well. Today, shared public utilities give us access to clean water by simply turning on the tap; cloud computing works in a similar fashion

–Vivek Kundra, Federal CIO, United States Government, 2010.

It’s not that cloud architecture is ground breaking and has not been tried earlier. But many things lined up for cloud architecture so it could become the galloping horse it is today. Cheap, reliable bandwidth and compute. Massive improvements in decoupled storage (scale/performance) and main-streaming of unstructured data storage and retrieval, allowing for complex analysis of data to be done over a programmable and elastic infrastructure has opened up the avenues where venturing out before was improbable.

Locked to a cloud or agnostic of cloud?

AWS, Azure and Google Cloud have captured a big share of the cloud platform space due to the fact that they provide the most number of lego blocks in the assembly for the end to end application. They lock down customers to their cloud by providing higher order managed services, where the complexity of these highly available services is managed internally by the provider. Managed services is the name of the game, the way to build and deploy applications fast. But companies wary of locking themselves to one cloud provider are overwhelmingly choosing Kubernetes, an open source application orchestration platform with it’s own abstractions for infrastructure. The good news is that AWS, Azure, IBM and others have started supporting Kubernetes as a managed service, Google cloud being a pioneer in this. This is a testimony to the exploding adoption of Kubernetes by the developer community. The popularity of Kubernetes can be gauged by the fact that docker, the company behind docker swarm has started supporting it. A few companies support Kubernetes for heterogeneous environments as well.

Open source is now pervasive

If Kubernetes is helping developers stay cloud agnostics, the maturing of several key technologies is making that decision even easier to take.

• Kafka, RabbitMQ, or both for messages and large events to be handled at scale.
• Cassandra, Elastic, Postgres – any or all depending on your needs to store data.
• Elastic + Fluentd + Kibana – easy troubleshooting for day to day operations.

The advent of Kubernetes operators will provide these off the shelf deployment models for the above stacks making it more easier for user to build their software stacks.

Shortened development cycle

The confluence of maturing technologies is allowing developers to focus on customer use-cases and solving their problems. Most non-functional but equally important aspects, like Scale, HA, debuggability, taken care of by the technologies mentioned above. This has allowed enterprise grade software to be ready in a very short time.

Right Shift -> Mobile first, Cloud first to Machine Learning First

Microsoft - Our strategy is to build best-in-class platforms and productivity services for an intelligent cloud and an intelligent edge infused with artificial intelligence (“AI”).

Google - Machine learning is a core, transformative way by which we’re rethinking how we’re doing everything.

Amazon - The most exciting thing that we’re working on in machine learning, is that we are determined, through Amazon Web Services, to make these advanced techniques accessible to every organization

In the current cloud computing space dominated by Google, Microsoft and Amazon, one technology is disrupting the current status. It is Machine learning. It’s no surprise that these companies now have a “ML first” strategy for their future. ML lives and flourishes on data and they have a boatload of data that they can use to hone their ML algorithms.

With their current market share, is there any scope for new entrants in the field of Machine learning? For sure there is. We will see more and more companies solving very specific problems using ML algorithms. Medical diagnosis is one such example.

Conclusion- Where will this all settle down?

The coming years will see more and more companies provide ML based insights for specific vertical use cases of customers. Algorithms and learning models as a service will be an API to be called. Companies buying these services will insist on greater prediction accuracy. This cannot happen unless customers agree to let their data be used by these service providers. We will see new contracts being drawn up to allow the use of customer data for improvements in prediction accuracy (which goes against data privacy, but that’s another discussion for another time).

Customers will also see more and more of their data scattered across such service providers, probably across clouds as well. Its logical that these customers will demand insights from their own data spread across clouds residing with different service providers. This should lead to new categories of applications working across clouds to gain insights for customers.

So get ready for the future where ML based prediction services are available for very specific customer needs, applications that stitch these services across clouds. Applications helping companies make sense of their increasingly complex data.

There is space for all the established leaders, the niche players and the consulting companies.

p.s. Many thanks to my friends Vibhu Pratap,  Atharva Chauthaiwale and Mahesh Kulkarni for helping me refine my thoughts.