We dove deep into our blog archive and discovered a gem from 2015, "Cloud Computing in 2020: Looking into that Crystal Ball." We got a few things right and we got a few things wrong, including the circa 2015 concept of a cloud architect being an "emerging idea." I thought it would be great to write an update on where we are now with these former predictions. We asked David Jones, GreenPages' Director of CSP Operations, to go back to the future and respond to our 2015 predictions. Read his responses (in green) below! Let's just hope Biff doesn't get his hands on this article in the past, you never know what he could do with it!
Cloud Computing in 2020
1. A utility and computing grid
In 5 years, large companies will still hang on to their datacenters to run some services. However, with security more robust, I think that corporations will make available their own computing resources as much as they consume cloud resources – just like some households generate their own electricity and sell it back to the grid. I think Cisco's Intercloud concept has an angle on this.
Clearly, large companies (all companies, aside from born-in-the-cloud startups) are retaining their datacenters. The capital investments are too great to discard before their end of life. However, I am not aware of any companies aside from the public cloud providers (Amazon, Microsoft, Google, Alibaba, etc.) selling their excess capacity back to the public. Looking forward another five years, I don’t really see this happening.
2. Flexible capacity, consistent access, and high portability
A cloud/compute socket just like an electrical socket. Standardized applications and connectors that “plug in” to the grid and are removed just as easily. Virtualization has the first stab at this, encapsulating the OS, data, and applications neatly in a VMX and VMDKs. Containers are the next stab. Redhat has an angle on this with their CloudForms PaaS. Raw compute power becomes more and more of a commodity as portability improves; meaning downward pressure on IaaS prices will remain to some degree (see #4).
We have not made the progress we should have on standardizing applications, at least when it comes to virtual machines. VMware, AWS, Microsoft all have their proprietary VM format. They can be “converted” between providers using the appropriate migration tools, but they are not what I would consider pluggable. Conversely, containers have made huge strides in this area, but there still doesn’t seem to be a universal container image format.
3. IoT or machine-to-machine communications/transactions
One machine determines that it needs to acquire more compute power to complete its work. It makes a "deal" to go out and acquire that compute power, uses it, and gives it back to the grid. Or, on the flip side, a machine that knows when it can stand idle and rent its own power. Another angle on this, a virtual machine or application has knowledge of its SLA, and moves to the provider who can deliver on that SLA at the least cost. Love it or hate it, Apple's Siri has an early angle on this. From what I've read about the technology, queries to Siri find their way back to Apple datacenters, not only to obtain answers, but to improve the accuracy of queries for all Siri users.
This is happening today if applications are architected and built to take advantage of the technology. Auto Scaling, Elastic Load Balancing, etc. provide applications with the ability to scale in and out based on demand. Combine that with concepts like spot-instance pricing, and you have the ability to provide compute power on demand that releases back to “the grid” when the work is complete.
4. Superesilient applications
As prices for cloud trend downward and portability improves (see #2 and #5), disaster recovery will take a new shape. Instead of running on a 2-site/2-region DR architecture, applications will run on a 5, 10, 20, or 30-site “DR” architecture, with all nodes being active. Does it matter where your application is running at that point? Potentially, it’s running all over the east coast, or all over the country. Some services from AWS already have an angle on this with services that are redundant across regions (a.g., S3, elastic load balancing, etc.), not to mention things like DNS on the Internet. I think it will become cost-effective to do this, in general, within 5 years.
This has absolutely come to fruition. Within a single region, technologies mentioned above (Auto Scaling, Elastic Load Balancing, etc.) provide single-region redundancy, while enhanced DNS services like AWS Route 53, Azure DNS and Azure Traffic Manager, AWS CloudFront, and Azure Content Delivery Manager, and inter-region replication for database and storage provide the ability to run truly globally redundant and super-resilient applications.
5. Compute traded as a commodity, just like crude oil
This might be a stretch in 5 years, but with the trend of IaaS being more commoditized and portability improving, we'll see a day when compute power is traded in a commodities market. In the channel, this is already fairly common – IaaS providers are eager to cut favorable deals with resellers who agree to purchase large chunks of infrastructure upfront, only to resell at a later date.
I remember the pipe-dream around this circa-2015. I seem to recall VMware was pushing cross-cloud vMotion and an integration with the ITBM (now vRealize Business) product that would identify the lowest cost public cloud provider and live migrate your workload to save money. I’m not aware of any realization of this.
6. IT and the business coming together
DevOps was the first marriage of two groups that had been previously at odds (oftentimes). Within 5 years, I think maturity in IT will improve to the point that they become as focused on the business as any other traditional LOB. IT becomes an Innovation Center -- they are focused on the business, and behave proactively. Corporate IT shifts its focus from requirements to possibilities. See my previous posts on the emerging idea of a cloud architect who will be important in this shift.
I agree with this; IT is definitely aligning better with the business. Customers are moving traditional IT systems that require a lot of care and attention, such as Exchange, to Office 365 so they can focus on more strategic initiatives rather than tactical systems administration. As companies mature their public cloud policies and governance, it gives consumers of public cloud (app owners, developers, etc.) direct access to the resources they need to accomplish their goals, without incurring massive monthly bills. IT is now able to focus on innovation and the business itself, rather than individual applications.