Cloud isn’t a one-size-fits-all solution. Hybrid infrastructure is shaping smarter decisions for cost, control, and compliance.
Mark, as the Chief Technology Officer at Yellowbrick Data, can you walk us through your journey in the data management and cloud technology industry and what has shaped your vision for Yellowbrick Data?
I’ve been the CTO at Yellowbrick for the past 5 years. Prior to that, I was a vice president at Teradata responsible for the data warehousing product line and Teradata’s ecosystem products. I was also Chief Architect of RainStor, a data warehouse archiving product that Teradata acquired in 2015. Before Teradata, I was at RainStor for ten years. I’ve been in the data management and cloud tech industry for 20 years now. I first got into cloud technology in 2010, when we ported RainStor to AWS and Hadoop to take advantage of compute elasticity and separated object storage. RainStor’s SQL query engine was forerunner of many of today’s modern data platform architectures. At Yellowbrick, we also use the same elastic and separated compute and storage pattern in the cloud and on-premises but enabled with Kubernetes.
What do you think has caused the noticeable shift towards “cloud-to-on-premises repatriation” and how does it impact organizations’ decision-making processes?
I think the trend of moving certain workloads from cloud to on-premises data center is a refinement as organizations mature their cloud usage, rather than a full-scale reversal of cloud adoption. The attractiveness of cloud has always been due to elasticity and on-demand consumption and billing. For long-running workloads, it’s often more cost effective to run them on your own hardware. More recent developments in open-source object storage software and in Kubernetes, also means you build a data platform with a cloud-like user experience in your own data center.
How have rising cloud costs influenced businesses to reconsider their cloud strategies and explore alternative on-premises solutions?
There have been well-publicized articles published by 37Signals and Andreessen Horowitz that talk about 30-50% cost savings when repatriating workloads to on-premises. I certainly think this is true for workloads that are ‘always on’. However, I believe the reality is that organizations are adopting a hybrid cloud strategy to reduce cloud spend, where occasionally running workloads remain in the cloud, and constant workloads are moved to on-prem. There are many more factors influencing the decision to repatriate workloads than cost alone.
Security concerns seem to be driving more companies to bring data back in-house. What are the most critical security factors prompting this shift, and how do organizations manage them effectively?
The security breaches in cloud data platforms that have occurred over the past couple of years have certainly influenced the thinking in some organizations. In fact, one of our customers is reconsidering their move to the cloud because of cloud service provider-level security concerns and want to reduce risk with a hybrid approach where they can divert users to an on-premises Yellowbrick replica in the event of an attack on a CSP. In financial services, cloud concentration risk is a concerning issue. Hybrid and multi-cloud solutions provide the flexibility to spread the institutional and systemic risks from having critical systems in a single cloud or data center.
Data sovereignty is becoming increasingly important for various sectors. How can businesses in regulated industries ensure they comply with local data laws while still leveraging the benefits of cloud and on-premises environments?
Countries are implementing strict data sovereignty laws that require data to remain within national borders (e.g., China, India, Europe). These laws are proliferating rapidly, with an expectation that more countries will adopt similar regulations. Non-compliance with data sovereignty laws can lead to fines, legal actions, and reputational damage. Some countries have even weaponized tax rules to try to get what they want in terms adherence to data sovereignty laws.
Often a hybrid cloud approach is required to address data sovereignty. In some instances, the public cloud provider doesn’t have a point of presence in certain countries — in parts of the Middle East and Africa, for example, and so the only way to ensure data is retained in-country is to run it in your own data center or in a colocation facility. Deploying a Kubernetes-based data platform on-prem and in country can provide many of the benefits of cloud while adhering to local data laws.
With the growing emphasis on reducing cloud reliance, how do companies balance between maintaining security and compliance while achieving operational efficiency through their IT infrastructure?
A key step is to avoid the proliferation of many different technology stacks across the organization, both in the cloud and on-premises. Different stacks imply different skillsets and hiring challenges, and more expense through software licensing. Not to mention increasing the complexity of the data management ecosystem. Companies should aim to standardize on a data platform that provides the same user experience and management skills wherever it is deployed. Tech stacks built on Kubernetes can help here. Kubernetes and container-based software is agnostic to the platform it runs on. Kubernetes can also provide compute elasticity, and the resilience needed for business-critical operations.
The decline in hardware costs has made on-premises infrastructure more appealing. How has this trend influenced the strategic decisions of organizations when it comes to repatriating workloads?
It’s not only the decline in hardware costs that is influencing the move repatriate workloads. It is also the rise of the private cloud, enabled by Kubernetes and cheap object storage, that is causing a rethink. Now, a data platform running on a private cloud can provide many of the capabilities available in public clouds, but with a lower total cost of ownership, particularly for constantly running workloads. Further, these on-prem data platforms can provide compute elasticity within the constraints of the hardware with a more reliable performance SLA for latency-sensitive applications. Kubernetes-based data platforms make expansion easy too: you can add new hardware to the private cloud server-by-server without disrupting operations.
In your view, what are the most significant benefits of bringing workloads back on-premises versus maintaining a full cloud infrastructure?
The most significant benefits of bringing workloads back to on-prem environments are: cost optimization, particularly for steady-state workloads; performance and control over infrastructure for low-latency applications; and data sovereignty and security for sensitive workloads. However, many of our customers are investing in a hybrid cloud strategy, where they get the best of both worlds, optimizing the placement of workloads based on business needs.
As businesses increasingly choose to repatriate data and workloads, how do you see the future of hybrid cloud environments evolving? What role does Yellowbrick Data play in supporting this transition?
In the past hybrid cloud environments have been difficult to set up and maintain, because they would often be based on technologies from different vendors. This caused the tech stack proliferation and skillset difficulties raised earlier. Yellowbrick helps here by delivering a SQL data platform that runs the same anywhere. It’s the same database software deployed in the public cloud or on-premises. Also customers can replicate databases to and from public cloud to on-premises to support disaster recovery and data movement use cases.
Looking ahead, how can companies ensure a smooth transition back to on-premises or private cloud setups, especially when managing complex workloads and critical data? What advice would you give to organizations navigating this shift?
Obviously, enterprises should identify the workloads that will benefit most from a move back to on-premises and build a business case around these. Once the business case for moving workloads has been validated, then look for data platforms that provide uniformity of operation and user experience wherever they are deployed. When evaluating workload repatriation, it’s often a good time to think about how to modernize the data platform too, in order to reduce the number of different data management and analytics technologies in play.
Quote:
There’s an addage from ten years ago that is becoming more relevant with the growing trend of repatriating workloads to on-premises: “Own the base and rent the spike.” In other words, use public cloud when you need agility and on-demand consumption, and target steady-state workloads as candidates to move back to your own data center.
Mark Cusack
CTO at Yellowbrick
Mark Cusack is the CTO at Yellowbrick. Mark has worked in the data warehousing and advanced analytics space for the past 20 years. He was a co-founding developer at the data warehouse archiving company RainStor, which was acquired by Teradata in 2014. Mark holds a PhD in computational physics, and has worked in academia, government, startups and enterprises over the course of his career.
Headquartered in Mountain View, CA, Yellowbrick Data was founded in 2014 on the principles that data is a critical asset. It must be secure, available to the right people at the right time, with resources prioritized according to organizational role. Above all, the data entrusted to you by your customers must be guarded.
Our SQL data platform does not expose data to the public internet. It operates in a decentralized manner, within your own cloud accounts and on-premises. Data is fully secured and protected according to enterprise policies.