How to Converse in Cloud: Precipitation vs. Repatriation
Recent personal engagements with analysts and my reading of a future-looking study got me ruminating about some often-cloudy weather I encountered in my way-back rainy days of college. For some reason, I found myself dwelling on Ernestine, a long-time cashier at the university student cafeteria.
Ernestine brought the positivity of Ted Lasso to brighten the spirits of every student standing in her checkout queue. When it rained (and it rained a lot in my college town) Ernestine would always say “rain is just liquid sunshine, baby.” That stuck with me all these years and just recently bubbled up to the surface while I was considering how my modern world of high-tech clouds is regularly faced with changes that some might consider as stormy times.
Cloud precipitation: Managing hybrid multicloud migration holistically
The cloud topic that got me ruminating about those long-ago rainy days is called “cloud repatriation.” That’s been a hot button subject in cloud conversations since Andreesen Horowitiz posted an analysis entitled “The Cost of Cloud, a Trillion Dollar Paradox”i — and it’s something that came up in several conversations during our Analyst Day event in June.
Of note is the ambiguity of the term repatriation. For many in the IT industry, the term has been used to describe the failure of overly aggressive, usually lift and shift, cloud migrations. Legacy application architectures often see lower availability, poor performance and high cost when directly migrated into the cloud — leading to a return of those IT functions back to legacy deployments (repatriation).
What authors Sarah Wang and Martin Casado describe in their Andreesen Horowitz analysis is something completely different. They focus on the cost optimisation of cloud native applications at scale associated with building, implementing and managing infrastructure more akin to a private hybrid cloud. I like to think of this kind of hybrid evolution involving a kind of functional cloud precipitation.
At Equinix, we’ve had a front row seat to the “post-cloud” precipitation of essential components out of the cloud and into hybrid architectures. One of the earliest examples of precipitation was the development of Open Connectii by Netflix. Content delivery network (CDN) as a service (or cloud CDN) was an essential part of the streaming giant’s early strategy, but content delivery became a significant portion of their cost of operations and a highly visible portion of their infrastructure. From a business perspective, it made sense for them to own this highly visible component of their value chain. As a result, Netflix built its own best-of-breed CDNiii — a CDN optimised for its massive streaming environment.
Think of precipitation as being a process that follows the water cycle. In literal meteorological precipitation, a fluffy white cloud accumulates condensation and becomes full before finally leading to rainfall. In the case of a virtual cloud, that “fluffy white cloud” is the beginning of an application of service in the cloud.
Choosing a single cloud vendor, at the infrastructure-as-a service (IaaS) and/or platform-as-a-service (PaaS) layer, gives an enterprise the flexibility to build scalable applications without consideration to infrastructure. As enterprise applications grow and scale, public-cloud providers offer a platform to meet infrastructure scale and distribution needs. When an enterprise needs global deployments, most cloud vendors make that simple and on-demand capacity aligns spending to usage. This model works well initially, but there comes a tipping point where the economics of scale require an enterprise to reevaluate the infrastructure design.
As we know from meteorology, when a saturation point is reached, some water will naturally precipitate out in the form of rain. As with a rain cloud, not all the “water” comes out in our virtual precipitation scenario. Companies must find a balance between public cloud and private infrastructure that allows them to optimise the economics of specific applications into a hybrid architecture. There is no single design that works for all applications. Application architects must evaluate the cost components of each part of an application. As illustrated in the Andreesen Horowitz study, for Dropbox, this meant building its own storage infrastructure. For others, optimisation has been around data egress, application performance due to placement, or the ability to leverage specialised processing hardware.
Many pundits claim that these kinds of examples are isolated cases and not indicators of a precipitation trend. What’s lost in these conversations is the well-established prediction that all business must become fully digital to survive. We’ve certainly seen a clear split in the retail sector. Those companies that have been able to make aggressive digital transformations and leverage the digital plus brick and mortar approach have survived.
Now let’s peer into a hypothetical crystal ball. Assume that transformative digital leaders are the only ones to survive into the all-digital future. If that’s the case, then how would a significant portion of those survivors not reach a scale and maturity that required accumulative precipitation of core digital components to remain competitive?
As a final point, the comparison of cloud environments to public utilities, while somewhat analogous, is not entirely accurate. The proliferation of features and services in the IaaS and PaaS layer is a strong indication that — in contrast to generating and consuming electrical power — there are nuanced and specialised components associated with cloud infrastructure. You don’t see that kind of capability proliferation with public utility power generation. That said, much like their power-grid counterparts, cloud providers have done a great job of meeting market needs while leveraging scale to keep costs low and maintain margins. Nonetheless, the specifics of the cloud business model highlight the fact that, for consumers at scale, there are still considerable opportunities for technology and cost optimisation.
No single forecast can predict the future
Not all clouds will result in precipitation (either meteorologically or virtually), and even those that do have varying degrees of rainfall. In the digital realm, people must start thinking of precipitation as a liquid sunshine. Rain is vital to a healthy environment as is digital precipitation. Application architects should expect and plan for rainy days. At Equinix, we provide a global interconnection and data center platform to optimise digital precipitation.
Learn how organisations can accelerate, enhance and ultimately benefit from a digital infrastructure strategy that is anchored in hybrid infrastructure and connected ecosystems, read the Equinix-sponsored IDC Analyst Report entitled “IDC — A Blueprint for DX Success — Start with Hybrid Infrastructure and Connected Ecosystems”.
i Andreesen Horowitz, “The Cost of Cloud, a Trillion Dollar Paradox,” Sarah Wang, Martin Casado. May 27, 2021.
ii Netflix, Open Connect.
iii GigaOm, blog post, “Forget the CDN players, Netflix is caching its own video,” Kevin Fitchard, June 12, 2012.
Balancing risk with agility for better government services.
The rapidly changing landscape within the industrial automation sector has prompted AVEVA to...
How to build and connect digital infrastructure with hybrid multicloud at the edge.