Sunday, December 28, 2014

Looking back on my 2014 Cloud Predictions

Here's my self-review of my predictions made one year ago:

=-=-=-=-=-=-=-=-=-=-=-

1. Amazon Web Services will continue to dominate as the leading provider of infrastructure-on-demand services. Like 2013, the number of new offerings will be smaller than previous years as much of the low-hanging fruit has been picked. More emphasis will go into stabilization and features to enable highly reliable computing (across regions).
>> Correct, but this one was too easy. 

2. In 2014, we will see OpenStack adoption in the enterprise in a more significant way. This will be both a blessing and a curse. Corporate Infrastructure & Operations teams will be challenged to create stable solutions around OpenStack. The premature release of poor products/code like Ceilometer, Heat and even Neutron will cause unnecessary pain. I&O teams will retreat to the most basic functionality. Those organizations who chose to use the open source bits and not license a commercial product based on OpenStack will rethink their decision. Issues related to a lack of quality and product management will continue to plague OpenStack throughout 2014, leaving the door wide open to VMware.
>> Partially correct. IMHO, OpenStack saw more adoption by service providers while traditional enterprise dipped their toes in the water. And yes, the door was left wide open to VMware.

3. Organizations that were unhappy with the VMware software licensing fees remain unhappy but begin to see it as the 'stable solution'.
>> Correct. 

4. The excitement around containers continues to grow in 2013. The Docker model moves beyond early adopters into early majority for dev/test workloads. The way in which Chef/Puppet are used shifts from run-time stack creation to design-time creation, followed by image snapshots. This extends the emphasis on 'idempotent infrastructure'.
>> Correct. 

5. As 'idempotent infrastructure' for individual machines begins to feel like a solved problem, the focus shifts to multi-tiered, complex application architectures. A new emphasis is placed on orchestrating full stack solutions regardless of cloud API, hypervisor, operating system and configuration management tool. On-demand, full stack provisioning across dev-to-prod environments becomes doable by your average Joe. That said, TOSCA fails to get any traction due to unnecessary complexity, lack of practical applications and are beat to market by light-weight open source solutions.
>> Correct. We saw a bunch of Docker orchestrators emerge; perhaps too many. 

6. More aggressive organizations implement resilience features into their cloud architectures. Systems move beyond the traditional auto-scale and adopt patterns to auto-heal tiers via closed loop monitors that re-provision failed systems and tiers implement run-time service discovery (to reconnect the topology).
>> #FAIL. This was wishful thinking on my part. I need to quit doing that. 

7. As resilient architecture become more common, so does the need to test them. Variation of the Simian Army enter the Global 2000; instrumented systems capture data key metrics (MTTR, data loss, etc.) enabling architects to improve the integrity and availability of their solutions.
>> Again, #FAIL.

8. The Hybrid Grid is born. Unlike prior grid emphasis, the focus is on long-running services (not batch jobs). Unlike Hybrid Clouds, the focus is on uniform containers (think LXC/Docker) and resource schedulers that are "service first" not "machine first" enabling the grid controller to offer reputable Service Level Agreements. In 2014, Hybrid Grids gain traction in the high-tech web-scale shops but remain out of reach for most large enterprises. A few small service providers will begin to offer Hybrid Grid as-a-Service.
>> Partially correct. The rise of containers on Amazon and Google Kubernetes or managed containers is pushing this forward.


9. As we closed out 2013, the Target credit card breach was front-page news. Concerns around security and compliance are increasing in virtually every industry. Removing manual processes is seen as an easy way to implement tighter compliance and security. In 2014, companies will implement sandboxes within their cloud that focus on specific problems like PCI compliance. The DevOps processes will mandate that software 'run the gauntlet' (network segmentation, anti-virus, code inspections, encryption, etc.) to provide a safer environment. More security professionals will singing the praises of the cloud as a means to automate inspections, lock-down environments and provide audit trails.
>> #FAIL. I was thinking we'd see much more SecOps or RuggedOps - but this movement is still early. 

10. I anticipate that 2014 will be the year of "cloud in a box", where converged infrastructure solutions and cloud stacks are pre-packaged into turn-key rack/row clouds. Stripped down versions of OpenStack will be the preferred controller. We should expect the usual hardware vendors (Dell, HP, Cisco, EMC, etc.) to offer their own brand, as well as to offer hybrid solutions that leverage hardened cloud software from Red Hat, Canonical and others. Companies that once looked at vBlock or FlexPod open up their wallet. Organizations that avoided converged infrastructure continue to avoid it - but create their own reference architectures for their home-grown kits.
>> #FAIL. Damn. This should have happened by now ... maybe one of my friends at HP, Dell or IBM can comment on why it's taking so long. 

=-=-=-
Alright - I didn't do so hot in my 2014 predictions. Most of the stuff I still agree with - but we'll likely have to wait for 2015 or 2016. Here's to being patient!
Jeff

No comments: