Skip to content
Article

The Cloud is Sh.., Long Live the Cloud!

The voyage to a new land… full of hope and promise and wonder. Flyers are handed out at the docks to entice anyone brave enough to make the journey. The flyers themselves emphasize the hopes and promises of a better life:

  • A shift from having to manage your own devices. A pay-as-you-go model, no more need to put upfront Capex.
  • Ready built “platform as a service offering”, an opportunity for customers to focus on their code, and not on infrastructure plumbing.
  • Instantly scalable and geographically distributable.
  • And of course, how can anyone possibly even imagine getting involved with GenAI or Data Science per se, without “clouded” infrastructure?

A promised land full of wonders and opportunities was presented with a whole host of new features and everyone was enamoured. And why not… it all sounded perfectly reasonable. Spectacular treasures to discover and a whole host of new plants, spices, and animal life to gaze upon. And of course, our cloud-native forebears whom we meet upon arrival who seem so much more advanced and savvy.

And thus… many began the journey…

The Journey2Cloud

Many organizations faced with cloud migration, normally make quick strategic decisions at an organizational level to prepare for cloud migration to rebase their financial spend on cloud technology. Namely, to reduce CapEx for setting up or renting data centers, purchasing servers, and hiring personnel to manage everything. The decisions were fast and sweeping and with such a sense of urgency everyone felt they needed to be “on the cloud”. Being on the cloud was the beautiful intersection of cost-saving and innovation (win-win).

What followed was the unfortunate small print of the flyers that sent everyone chasing new lands!

The first challenge… the cloud is slow(er)

Cloud migration… Why not, there is even a simple Journey2Cloud process to follow the first step being to “simply” lift and shift your existing VMs or physical machines onto the cloud infrastructure.

This seemed almost foolproof, but once the migrations started to take place, so did the cloud consumption costs. This was in part because of the pricing models of the cloud providers and in part down to the small print that says cloud infrastructure is slower than physical infrastructure. And thus once on the cloud, the only way to increase its performance to pre-cloud days is to scale it.

Hence Lift and Shift tends to cost more than anyone first anticipates, namely because you consume more than you anticipated… The cloud world is slower than the physical world and the only means to make the difference is to rent more resources.

Quick note: to highlight this fact, if you were to measure the time taken for packets to travel between New York and London, you would find with data centers that this would be more or less the speed of light. But with cloud infrastructure, it would be in the order of 10 times slower.*

The Second Challenge… the Performance Paradox of Hybrid Clouds

Any long-standing non-cloud native organization undoubtedly has an IT architecture landscape with lots of valuable and important databases/data sources that are mission-critical and of immense value to the organization. So it stands to reason… the first few migrations and pilot initiatives will tend to evolve moving web services or application processes to the cloud whilst connecting to the (valuable) data in the existing on-prem infrastructure. Hence we kick off with Hybrid Cloud.

It makes so much sense. Reduce complexity and risk of having to worry about affecting existing systems or processes, and peel off areas that can be rewritten or migrated to the cloud in atomic pieces. However, this tends to introduce even more performance issues in the form of latency. The workaround for this is to scale horizontally, which further increases consumption and hence costs.

Migration of core data to the cloud is often the very last step in the game and for good reason. Transactional data is most often the heart of a business and it’s a most valuable asset. A better approach to waiting with the core transactional data until the last minute would be:

  • Moving data is better tackled earlier
  • Scaling horizontally requires refactoring
  • Legacy apps can also scale vertically (bigger boxes)
  • Finding ways to make legacy applications less chatty (make latency less of an issue)

The Third Challenge… Somebody Call FinOps

Before the cloud, infrastructure was considered to simply be an investment (Capex) already spent, and thus the name of the game was to simply consume consume consume. But now everything is running on a meter. VM machines burn more consumption than Serverless components, processes suddenly get scrutinized for how they can be changed to work on demand. Enter Financial Cost Management.

If you ever wanted to be distracted from your core business and trying to innovate new services or products, Financial Cost Management could be regarded as a weapon of mass destruction to that end. Enter FinOps…

FinOps is an operational framework, cultural practice, and most often also an organization that optimizes cloud cost by engaging cross-functional teams. FinOps becomes responsible for understanding how and where to tackle technical debt vs delivering new features to the consumers of the platforms. Like DevOps or Lean Management before it, the best approaches here tend to revolve around making problems visible (i.e. a bunch of big monitors with dashboards) so responsible stakeholders can meet regularly to agree and set priorities.

Financial Cost Management is a rocky phase that is best managed and sailed through with support from the senior executives of an organization to avoid it becoming too turbulent. It is a period where unwavering determination is needed to come through to the other side (i.e. a cloud native-like architecture) whilst not disturbing planned product features and releases of existing systems. And new organizational setups to meet, discuss, plan, and agree cloud decisions and establish and evolve cloud best practices and also mature business services to tag how much cloud spend they cost and hence how they should be priced to customers.

Rejoice at Landing on New Shores: The Land of the Cloud Natives

Anyone coming through the challenging journey and setting their feet down on this new land can rejoice at the fact that they made it! And what a strangely exotic land to encounter. The soil is rich in all sorts of new building blocks for applications, ranging from API management tooling, a whole host of different data storage and warehouse possibilities, caching, monitoring, elastic scaling, and of course lots of clever machine learning capable tooling.

Empowering and supporting rapid application development and innovation. Meanwhile, despite the challenges, the organization itself too should have gone through a digital maturing process that leaves it more robust to think digital. Not just in technology, but in culture, mindset, and processes. New ideas and prototypes can be spun up almost instantly, with a fail/succeed fast attitude, services can also be more readily spun up seasonally for a special event or holiday, and computing resources can be scaled horizontally elastically to account for high peak demand.

Computing consumption gets easier to meter and the pricing of services can evolve to account for end-to-end costs, bringing greater transparency and data to organization leaders, and enabling organizations to rethink their go-to-market strategies for their services.

Infrastructure that can be initiated with code. Releases automated. Resources can be spun up just in time to manage large batch data analytics/AI tasks. Architectures become easier and faster to adapt.

The wheels of innovation begin to turn quickly. Change begets more change, creativity drives more innovation. And the cloud becomes the catalyst for the faster evolution of digital products and services. Once again, not just for technology, but for the organization as a whole as it awakens to leveraging digitalization across the organization.

The Future of Cloud. Adoption Is Everything!

I have a dream… I have a dream that in the future all IT infrastructure will be cloud-based and organisations will no longer have to be concerned with the management of said infrastructure. More rather, they will be free to focus on their core business(es).

I have a dream… that the cloud will be a great enabler for even more creativity, innovation, and faster evolution of digital products and services. With cutting-edge technologies ranging from cryptography, GenAI, or ready built self-management tools all embedded into the system. So what is holding us back… the handbrakes…

Adapting Organisations for the Cloud

The Journey to the cloud is not without its challenges as discussed and many organizations need to consider the wider changes it may need to their teams and departments to manage the migration. Which can be a handbrake for some. Transforming organizations, planning migrations of legacy systems, and hiring personnel to manage what were traditional organizations a few steps closer to cloud-native organizations takes time change management i.e. time.

Privacy-Preserving Technology

Whereas for others, the case is more closely related to trust…

… Any business willing to put all of it’s data in the cloud is aware that there are no absolute guarantees that no body or party will have the means to access it. Geopolitical issues are constantly swinging from one way to the the next, the business landscape where cloud provider cannot be excluded from creating competitor companies such as banks, insurance companies etc. a strong question of trust and security arrises. Not just from hackers or malicious attacks, but from the cloud providers themselves. How can companies be sure their secrets are truly safe?

With the advent of scalable homomorphic encryption and Trusted Execution Environments on the processors themselves. Perhaps the answer will be to interweave these principles into the cloud. Hence bringing confidential computing to the public cloud, and thus making it mathematically impossible (improbable) for anyone to ever be able to peek at someones data in the cloud, not even by the cloud provider themselves (or governmental intelligence agency).

A Greater Differentiator… Fear of Being Left Behind.. And More Cutting-Edge Features That Are More Present in the Cloud Than in Data Centers

The differentiator between cloud infrastructure and data center infrastructure is there, but the feeling of being left behind might not be big enough to draw more people. With an increasing amount of inbuilt services in cloud infrastructure for resource-intensive tasks such as Data Science/AI or GenAI. Perhaps this will be the point of mass migration, in fear of being left behind.

Cheaper Models for Running the Cloud

Of course, adoption and price go hand in hand. And whilst it may not be possible for most cloud providers to reduce the pricing of their services. Perhaps it might be more fitting to help provide more tooling and dashboards that support organizations with FinOps. At the moment, mature FinOp practices and tools are best brought into organizations via external experts. Perhaps one might dream a lot of the best in-bred tooling might become accessible and integrated into their respective cloud management portals.

Closing

Cloud technology is here to stay. Migration to the cloud for larger organizations is not always easy and they will be cursing about the cloud at various stages through the phases. But there are rewards and slowly it does present itself as a focal area that can innovate itself to provide new services such as GenAI for organizations and ease the process of them following technology trends.

  • LN/NY is about 26ms for light and 30ms in real life AWS US-East1 to our Iowa data center is 6ms for light and 60–70ms in real life).

Collaboration with:

  • Mike Josi
    Senior Director, SRE and CICD, Transamerica

Explore more articles