Podcast: Play in new window | Download (Duration: 34:17 — 39.2MB)
Subscribe: Apple Podcasts | Spotify | Amazon Music | RSS | More
Now that businesses have deployed modern applications in the cloud they are starting to ask whether it might be more attractive to run these on-premises. This episode of the On-Premise IT podcast features Jason Benedicic, Camberley Bates, and Ian Sanderson discussing the pros and cons of cloud repatriation with Stephen Foskett. A recent blog post by 37 Signals got the Tech Field Day delegates talking about the reality of running modern applications in enterprise-owned clouds, whether in the datacenter or co-located. Certainly the hardware and software are available to move applications on-prem, and some workloads may be better served this way. Most of the necessary components to run modern web applications are available on-prem, from Kubernetes to Postgres to Kafka, but these can prove difficult to manage, which is one of the things as-a-service customers are paying for. Looking back to the debut of OpenStack, enterprises have wanted to run applications in-house but they found it too difficult to manage. OpenShift is much more attractive thanks to the support and integration of the platform, but many customers have financial and administrative reasons for as-a-service deployment. It might not be a mass exodus, but there are plenty of examples of repatriation of modern applications.
Why Companies Are Moving Off of the Public Cloud
A new trend coming out of the enterprise IT industry is cloud repatriation. The chatter picked up when 37signals, a SaaS project management company, publicly announced that it saved $1 million by pulling apps away from public cloud. According to CTO, David Heinemeier Hansson, repatriation has shrunken the company’s cloud spend by 60%, and is projected to save an estimated $10 million over the next five years.
And theirs’ is not an isolated case. Skyrocketing costs of data and storage in the cloud have caused a lot of companies to pull away and migrate back to on premise datacenters in the last few years. Seagate has built its own platform to deploy web applications that runs in their private datacenter on owned hardware. More recently, LinkedIn has called off plans to move workloads from on-site to Azure Cloud.
So are companies really abandoning their cloud computing dreams and hauling wares back to where they started? At the recent On-Premise IT Podcast, host Stephen Foskett addressed this question that’s lately been the talk of Silicon Valley.
Public Cloud Offerings Come at a Premium
When considering relocating technology, the reasonings fall into two main buckets – cost and control. “As we went into 2024, a lot of very large enterprises are concerned about costs. So there is this ongoing effort for cost management, and what is happening is a recalculation or reevaluation of where the workloads are to be placed and why. That workload rationalization has been going on for some time,” notes Camberley Bates, VP Practice Lead at The Futurum Group.
Enterprises’ rationale behind migrating to cloud was to reduce OpEx. The cloud offered an attractive answer to the surging cost problem in on-premise datacenters. The promise however soured as companies started to struggle with cost blowouts. Despite adapting their operating principles and practices to rein in the spendings, an optimized cloud value has remained unrealized.
After expending a notably large amount of time and resources to get to cloud, when a company decides to withdraw, it reflects as poor planning. Much like in all financial decisions, the sunk-cost fallacy creeps in. And to keep the cloud obsession going, hyperscalers hook users in with free credits that give them a free pass to start down the road.
From Managed Cloud to a Private Infrastructure
Spurred by dependency fear and cost and ownership concerns, many big enterprises have started bringing selected applications on-site as part of their workload placement strategy.
“A few years ago, it was a cloud-first mentality which we’re moving away from today with the hybrid approach, but it’s a very interesting marketplace in terms of options of where you can repatriate to in terms of the software stack,” says Ian Sanderson, Product Manager.
One of the things that makes the argument of going back on-premises seem valid today is the evolution of datacenter computing. “Since the cloud came about, we’ve seen a lot of step change in on-premise compute. We have gone from average systems of 4 cores to up to systems with 64 cores. So you could pack a lot of compute into a small space at a small cost,” points out Jason Benedicic, independent consultant.
A growing technology ecosystem is making shifting applications possible. “There’s a lot more off-the-shelf products for running clouds. Kubernetes and containers have come a long way. So the skill ramp-up needed to build and run your own modern application stack is lessened – I don’t think it’s completely removed, it’s not as easy as virtualization is – but there’s a lower barrier to entry. There’s a cheaper, more dense hardware aspect and those come together to make repatriation a possibility,” he adds.
The Uncomfortable Truth
Although technological advances lend users freedom to place their workload anywhere that offers maximum cost, performance and security payoffs, lifting and shifting too has its trade-offs. The cloud has monopoly over a few things that enterprises can’t pass up on, especially with the wide-spread adoption of AI. For one, on-premise infrastructures can barely match the agility, speed and iteration of public cloud.
“If you run a startup business with a couple of DevOps engineers and a fairly small team, it is going to be a daunting proposition to run all of it yourself. It’s possible, but the question is, what are the hidden costs and where do they lie,” cautions Benedicic.
But, increasingly, data costs in the cloud are driving companies to rethink the strategy. “Talking about the issue of cost analysis, we’ve seen a decline in the cost of server instances. We have not seen that same kind of cost basis on the data side,” notes Bates.
An Emerging Ecosystem
Thankfully, modern containerized applications have some amount of portability built in. “With serverless stuff, there’s some level of interoperability but there are not a huge number of serverless platforms out there that are mainstream,” Benedicic says.
Companies like Red Hat and IBM have solutions that make quick work of installing on-prem environments. The rise of OpenShift has been game changing in the way people think about running private cloud. Red Hat OpenShift is an open-source container application platform. The on-premise PaaS flavor is self-managed and comes with on-prem support for maximum ease.
Red Hat is one of the companies that is building a full suite of tools that work together to make the transition easier. Things like deployment blueprints that serve as guides are extremely helpful to get users started.
Wrapping Up
Workload repatriation need not be a binary decision. In many big enterprises, cloud repatriation may have taken off, but it is not a quit-the-cloud movement that it has been made out to be. Amid an economic downturn, companies are trying to tighten their budget and deciding where a workload best resides is the cornerstone of that. A hybrid placement approach will ensure a more natural distribution of workloads across cloud and datacenters than we have seen before.
For more, be sure to check out the On-Premise IT Podcast episode – Cloud Repatriation Is Really Happening – to follow the discussion.
Podcast Information
- Stephen Foskett is the Publisher of Gestalt IT and Organizer of Tech Field Day. Find Stephen’s writing at GestaltIT.com and connect with him on LinkedIn and on Twitter at @SFoskett.
- Camberley Bates is the VP and Practice Lead at The Futurum Group. You can connect with Camberley on LinkedIn and her podcast Infrastructure Matters through The Futurum Group.
- Jason Benedicic is an expert in Kubernetes and Cloud Native Platform Engineering. You can connect with Jason on LinkedIn and on X/Twitter or read more on his blog.
- Ian Sanderson is the Product Manager – Data Protection at 11:11 Systems. You can connect with Ian on LinkedIn and on X/Twitter and read more on his website.
Read more about Cloud Repatriation and what some possibilities are: Our cloud exit has already yielded $1m/year in savings
Follow us on Twitter! AND SUBSCRIBE to our newsletter for more great coverage right in your inbox.
