Here are my seven predictions for what we will see in cloud computing in 2018.
1. The cloud’s dominance in business will slow
Don’t get me wrong, public cloud-hosted software-as-a-service will continue to eat traditional enterprise workloads. Visionary companies that recognized this trend, such as WorkDay, will continue to gain at the expense of enterprise software and even service companies like ADP. However, the torrid growth of the cloud will finally slow as forward-thinking businesses realize how to achieve cloud-like agility and efficiency and, most importantly, improved economics for their on-premises systems.
2. HCI will strike back
Hyperconverged infrastructure (HCI) will be the key to halting the massive migration to the cloud. A few years ago, IT administrators spent a huge amount of time and resources just to design, purchase and deploy the virtualized compute, storage and networking infrastructure required to support their required workloads. Enterprises had already burned through their budgets and burned out their IT admins — all before they’d even started developing the critical business-specific applications. No wonder CIOs have been moving everything to the cloud!
But HCI will change this in a major way in 2018. A single purchase order and HCI solutions like Nutanix and Simplivity will deliver to the loading dock a turnkey, private cloud that can rapidly be deployed. CIOs will suddenly notice how much they are paying for their “cheap” public clouds and realize that, with HCI, they can actually achieve the same agility, efficiency, scalability and improved cost compared to developing their apps on public clouds.
3. The network will become invisible, too
HCI has delivered on the promise to make compute, storage and virtualization infrastructure invisible and painless to deploy. Plug it in and it works. However, there is one little problem: HCI vendors have just been too busy to worry about the network, requiring IT administrators to still be networking experts in order to get everything working.
It turns out that 70% of all HCI support calls during deployment concern networking issues. The VMs and storage come up just fine and load-balancing and VM migration just work auto-magically. But problems arise when the network breaks everything because it is ignorant of all the automation going on at the compute, VM and storage levels.
I expect networking to finally become fully integrated with HCI platforms in 2018 and, as a result, make the network invisible, too.
4. Open networking will gain share at the expense of black-box incumbents
Open networking platforms will take an ever bigger slice of the networking pie. Customers have realized that the main purpose of closed systems is vendor lock-in and high maintenance and service fees. By contrast, open networking platforms disaggregate hardware and network software, giving customers the opportunity to choose the best in class of both.
5. Ethernet storage fabrics will hasten the death of Fibre Channel
Fibre Channel proponents (as they bounce their grandkids on their laps) will continue to shout: “I’m not dead yet.” And just like the plucky knight in the Monty Python film, Fibre Channel will continue to hang on, but lose another limb.
An ever smaller base of FC advocates will find new flash-centric storage technologies like NVMe-over-Fabrics a compelling reason to switch to Ethernet. And even the most adamant of FC proponents will be motivated to switch to an Ethernet Storage Fabric that delivers the reliability, familiar storage management capabilities and interfaces of traditional storage-area networks (SANs).
The demise of the dedicated SAN will be hastened because there is no Fibre Channel in the cloud, nor in software-defined storage, nor in HyperConverged solutions either.
6. Containers will eat VMs
Containers offer significant simplicity, portability and scalability advantages compared to deploying full-blown VMs in a virtualized environment. However, this has come at the expense of poor management tools and inferior isolation between workloads and shared network devices.
Nevertheless, containers have continued to evolve, and Kuberenetes has taken a clear lead as the container orchestration framework of choice. In addition, new APIs like Container Networking Interface (CNI) are solving the problems of security and network isolation.
Container success will come at the expense of VM management frameworks like OpenStack. OpenStack has failed to capture the enterprise as proponents had hoped, but it has found a solid niche in telco cloud deployments. Still, the OpenStack community is innovative and resilient and will pivot to embrace containers. Who knows, maybe they’ll hold the OpenStackTainer Summit soon.
7. Bitcoin will collapse in a fire fueled by its own success
Okay, I confess, this last prediction is only because I promised seven predictions and also because (ruefully I confess) I don’t own any Bitcoin. But I find the phenomena utterly fascinating and, as a fan of Neal Stephenson’s Cryptonomicon a decade ago, I should rightfully be sitting on a pile of #BitcoinBillions but, alas, I am not.
Gratefully, the big GPU-fueled clusters mining Bitcoin consume a massive amount of compute, network and storage. So, until the Bitcoin mining data-center operations self-immolate in a thermal run-away conflagration, at least technology providers will enjoy some portion of the riches fueling the Bitcoin billionaires.
-- Kevin Deierling (@TechseerKD) has served as Mellanox's VP of marketing since March 2013. He has more than 25 patents in areas including wireless communications, error correction, security, video compression and DNA sequencing.