Digital is about to break the Cloud - and it's a good thing

Fashion always comes back in a few years and, so does technology trend - as it seems. The centralized mainframes of the 1970s gave way to the de-centralized/distributed client-servers and 3-tier architecture of the 1990s and then we invented the Cloud - back to a centralized system. We have public, private, hybrid - all sorts of clouds in the enterprise and it's fair to say that Cloud Computing is driving today's digital transformation.

Enterprises now run 80% of their operations on the cloud and this pie is almost equally divided between private and public clouds. Gartner says "More than $1 trillion in IT spending will be directly or indirectly affected by the shift to cloud during the next five years".

All our digital services are powered by cloud - here is a snapshot of what happens in 1 minute on the internet (from 2016)

• 700k logins on Facebook
• 70k hours of Netflix
• 1.5k Uber rides
• 350k tweets in twitter
• 36k hours of Spotify
• 3million YouTube video views
• 125 new LinkedIn accounts

As these numbers sound very impressive - let's add the future use cases to the cloud:

• IIoT (Industrial IoT) will pump gigabytes of data into the cloud from every sensor
• Airplanes will collect and process a lot more data in-flight and become flying data centers
• Autonomous cars will create tremendous amount of data (and will need computing power of a super computer)
• M2M (Machine to machine) will grow exponentially with large data and computing
• Millions of algorithms and models will be created in the cloud that will crunch tons of big data

So, the obvious question is - will our clouds be able to handle this level of loads - both data and computing? Just to give an example - it is estimated that an autonomous car will create 6GB of data per kilometer. I think the cloud - as we know it, will struggle to inhale so much of data, the network pipes will struggle to move such heavy volumes and the processors will fail to crunch so many numbers. To me the most logical solution is - let's break the cloud and move things around. 

Cloud is our dumping ground

The move to cloud has been a win-win-win for all - enterprises, cloud providers, IT consultants - everyone has won. Enterprises saved millions of dollars, cloud providers raked in hosting fees and IT consultants increased billing rates - there is no reason for anyone to really complain. But, enterprises used the cloud as a dumping ground for everything - they dumped data in data lakes, shifted CPUs/GPUs to the cloud servers, made AI and ML models & algorithms on the cloud. Cloud is now the de facto always-on active archive. Moving to cloud has been the answer to every IT problem - almost like a Dilbert joke. I am not suggesting that these strategies were essentially wrong - it has worked so far and will work for many enterprises and services. But, it will break soon.

It's all about Latency

As digital transformations create new use cases - it will push the boundaries of cloud computing. In our simple use cases above (the examples of Facebook, YouTube etc. above) - we normally don't need to think about latency too much. When my Google search happens in the cloud and it takes a few seconds, or my Amazon order status takes a while to load or I get a bit of buffering in a YouTube video - none of these are critical in nature. But, the next wave of use cases are about to bring latency at the top of the list.

As drones becomes popular - real time data processing will rise drastically. As cars become autonomous and we move towards V2V interfaces - real time analysis will be the key. As billions of new machines join the IIoT network and M2M gains popularity - moving terabytes of data, crunching them and storing them will take a heavy toll on the cloud infra. We are now moving towards real-time processing of data - and nothing is more important than latency. When an autonomous car detects an exception - it will not have the time to send the feed to the cloud, have it analyze it and get back a brake message to stop in time. When a drone is flying with a missile and it detects an incoming hostile object - it will need near-zero latency. As a robot doctor performs a surgery and it detects a new pattern - it will have to run the ML algorithm immediately. I don't think today's cloud infra will be able to handle the requirement of zero latency.

Let's break the cloud

Let's analyze what the cloud will have to do now. It will primarily have 3 functions - almost like 3 tiers. Let's start with the outer tier:

1) Collect structured and unstructured data (from sensors, systems, environment): As I said, data collection will be huge - we will soon collect petabytes of data every minute and sending all of it to the cloud will be almost impossible. Hence the need is to collect and process the data close to its source. It will make a lot of sense to move this tier to the edge of the cloud and that will see the rise of Edge Computing.

2) Process the data (compute on the data, run algorithms/models): Next task is to process these data, which will include - extracting information from the data, finding patterns on the data, running defined algorithms on the data. There is a very good case to move these tasks to the edge as well. Depending on the use case some of these can probably stay at the traditional cloud. All the Peer-to-peer use cases like V2V and M2M will start taking some of these 2 tasks. Smart sensors will be able to consume and process data. In some cases the collect and process tiers can be combined into one and moved into the Edge Computing tier.

3) Learn from the data (ML and AI will learn from the data and re-tune the model): The 3rd task on the cloud will be to massage the data, store it and learn from it. In most cases this can afford a bit of latency and can be stored in what we now call a traditional cloud. This will eventually become the learning center of the enterprise. Machine learning, driven by AI will eventually push the learning back to the processing tier in the form of a new models and algorithms.

I do feel that we need to break the cloud to create a sustainable cloud platform. Edge computing is a very good start. If you look at the new cloud - which is broken into the layers as above - it does look like a 3-tier architecture. Oh, as I said in the beginning - technology is like fashion and it comes back in a decade or so.