GitOps and the Cheap Cloud Myth
Sarah Wang and Martin Casado of Andreessen Horowitz wrote a thought-provoking article on the fact that cloud vendors are taking a huge chunk of a company’s market value, leading many organizations to repatriate their infrastructure from cloud to on-premises. They conclude that there is an oligopoly that’s unsustainable in the long term. The three main public cloud providers will either start to give up part of their current 30% margins, or they’ll start to give up workloads. I’m no financial analyst for sure, so the element from that conclusion that interests me the most is workload delivery and workload repatriation.
Whatever the camp you may sit on, the myriad app delivery and continuous operations use cases currently being applied by any software company will likely push organizations to adopt a hybrid cloud approach: not fully vested on public cloud, not fully repatriated to on-premises machines or environments. Whichever the case, with hybrid and distributed approaches to compute, storage and networking comes environment sprawl that’s also known as utter chaos. Without control over infrastructure, organizations cannot effectively integrate and manage their application deployments, which hinders whatever value is encapsulated within their code. Without control over configuration, companies lose their developers and the chance to make a transformation and adapt. Finally, without control over workloads, companies miss the future altogether.
Let’s look at the reasons for adopting a hybrid cloud model, the challenges that come with it, and how best to navigate these uncharted waters.
The Cheap Cloud Myth
The cloud computing industry has grown very rapidly over the past few years. The consensus is that cloud computing is far more convenient, and sometimes affordable, than on-premises options. However, the reality is, as always, a bit more complex. Cloud computing costs are simply hard to predict and include a lot of shadow costs that eventually eat into your organization’s bottom line to the point which, like Sarah and Martin point out, they seriously detract from profit.
It’s easy and cheap to get started with the cloud, but it’s even easier to get lost in a sea of licensing costs that grow higher and higher, depending on your workload. For example, if you run a fairly static website using the cloud, you do not need to customize the server or OS. You can just wake up knowing that your website is up and running every day. However, this is not true for complex web apps or enterprise applications. They typically require myriad well-integrated cloud services, such as unique compute and memory requirements, different types of storage for different purposes, and a host of management tools across the entire software delivery process.
It’s a bit like owning a car versus renting an Uber. Sure, Uber is cheaper if you need just a few rides per month. However, most people can’t imagine life without a car. That’s why most still prefer the peace of mind and long-term savings of owning a car rather than relying exclusively on Uber.
The graphics below provide some interesting data points about the cloud spend of SaaS products.
Source: Harness
Source: Harness
The forecast of private clouds and data centers is spectacular. So it seems that, for a while, companies will be stuck with a hybrid setup.
Remember, cloud computing is really not as cheap as it seems. If you compare the hourly cost of a cloud instance to an expensive hardware server, it would seem that the cloud has an advantage. The public cloud’s promise of convenience and total uptime is still appealing, yes, but the growth of edge computing and networking, along with private cloud, just compound on the idea that hybrid is the way things are.
Security and Management Challenges
While shifting to a hybrid cloud environment, there are primarily two challenges: security and environment management.
Securing hybrid cloud environments is not easy. Hybrid clouds contain diverse infrastructure, which makes it difficult for security teams to monitor and protect. Malware is becoming extremely sophisticated, which means that security teams must adapt to new kinds of threats all the time. Enterprises are allocating more time and resources to security, but there are still vulnerabilities that result in data loss, downtime and reputation damage. Security protocols and controls have to be re-evaluated and implemented to ensure that all security aspects are taken care of.
One of the main concerns for IT departments when it comes to moving their data to the cloud is data sovereignty. Some companies are required to keep data within national borders, while others simply want to keep their data on their own servers. Whatever the reason, hybrid cloud is the only way to keep your data on-premises while still benefiting from cloud solutions.
Another hybrid cloud challenge is infrastructure and resource management. Organizations want to optimize their use of hybrid cloud environments and identify opportunities to move workloads across environments, and to do this without interrupting service delivery. Building automated processes is an additional goal. This allows them to spend less time on manual tasks and more time on strategic projects.
A Hybrid Cloud Management Solution
A hybrid cloud management tool should be environment-agnostic – equally functional and effective on-premises as it is in any cloud platform. One approach that has emerged as a strong candidate for hybrid cloud management is GitOps. GitOps is a relatively new (since 2017) approach to managing hybrid cloud infrastructure resources. GitOps is a set of practices that uses Git as the single source of truth to manage all parts of the system, from infrastructure to applications, from on-prem to cloud. It uses Git to store and manage configuration items. These items include not only the configuration files used to control the resources, but also the state of those resources and the events that have triggered a change in that state.
Weave GitOps is our solution, which combines the principles of GitOps as applied in leading open source tools such as Flux, Flagger and service mesh tools like Istio and Linkerd to enable hybrid cloud management that can span across both data center and cloud.
Conclusion
The cloud is touted as cost-effective, but lurking behind this story are shadow costs that can tip the balance in favor of on-premises. Organizations want to adopt hybrid cloud for regulatory reasons, to avoid vendor lock-in or to adopt best-of-breed tooling. Irrespective of the reasons, hybrid cloud management is not easy. The key is to have the right hybrid cloud management approach and solution. A solution that can cope with the differences across on-prem and cloud.