The theme of cloud repatriation has been getting some attention lately. While this has implications for software infrastructure companies, the effect will be varied. Mega-retailer Walmart has been on a journey over the last few years to modernize their technology stack. Their CTO unveiled another milestone in Walmart’s cloud native platform journey last week. They now claim to run one of the largest hybrid clouds in existence. This provides us with a good example of a company choosing to run some part of their cloud infrastructure themselves.

For their latest milestone, the CTO introduced Walmart’s “Triplet Model”, which represents three regional cloud deployments (West, Central, East) connected to 10,000 edge cloud nodes located in their facilities. This model brings computational power and data closer to customers and associates, resulting in better application performance and low latency. It also offers additional capacity for periods of peak consumer demand. Finally, Walmart claims they were able to reduce cloud costs by 10-18% annually using this model.

Within the Triplet Model, Walmart can deploy applications to both public and private clouds and seamlessly redirect traffic between them. This is enabled by an intelligent routing layer called Traffic Management, an open source cloud management platform called OneOps and the Walmart Cloud Native Platform (WCNP), which provides a cloud abstraction layer. WCNP currently manages a huge deployment of 545k pods on over 93k nodes.

Walmart’s modernization initiative was introduced in February 2020 during their Investment Community Meeting by the CEO and CTO. Walmart leadership acknowledged that the cloud was becoming central to a modern software operation, but they didn’t want to just “lift and shift” all of their existing applications to the cloud. Rather, they preferred to take advantage of what the cloud has to offer. They liked the idea of making all their applications cloud native, but maintain a control plane that determines where the applications actually run. The runtimes are distributed between three hosting models:

• Their large private cloud running in their own data centers (using OpenStack).
• Public cloud partners, specifically Azure and GCP (AWS is excluded presumably for competitive reasons)
• Edge locations hosted within their stores, bringing applications closer to customers and employees

Walmart anticipates three operational advantages to this model:

• Capacity management. They can accommodate temporary high traffic periods by spinning up more workloads on the public cloud.
• Use of best technologies. For each type of workload, they can harness the optimal solution available. In some cases, these are on the public clouds. Processing big data sets and training machine learning models represent one example. They also run much of their IoT data collection infrastructure through Azure.
• Improve performance by deploying some applications at their edge locations (generally within stores).

A question often raised by investors is whether this kind of use of hybrid cloud with a large private data center component represents a risk to software infrastructure companies. Like anything, it depends, and can’t be treated with a blanket assumption that it reduces demand. The distinction lies in the layer of the software stack and how much commoditization exists there. Also, the level of difficulty in running that software layer internally versus outsourcing the operation to system experts is a consideration. Let’s break this down into some examples.

---

Sponsored by Cestrian Capital Research

Cestrian Capital Research provides extensive investor education content, including a free stocks board focused on helping people become better investors, webinars covering market direction and deep dives on individual stocks in order to teach financial and technical analysis.

The Cestrian Tech Select newsletter delivers professional investment research on the technology sector, presented in an easy-to-use, down-to-earth style. Sign-up for the basic newsletter is free, with an option to subscribe for deeper coverage.

Software Stack Investing members can subscribe to the premium version of the newsletter with a 33% discount.

Cestrian Capital Research’s services are a great complement to Software Stack Investing, as they offer investor education and financial analysis that go beyond the scope of this blog. The Tech Select newsletter covers a broad range of technology companies with a deep focus on financial and chart analysis.

---

First, I think that the lower in the software infrastructure stack a component exists, the easier it would be to duplicate in a private data center. This encompasses infrastructure like the physical space, server hardware, network connectivity, operating systems and storage. At a large size, Walmart should be able to duplicate the compute, raw storage and network capabilities in their private data centers that could be provisioned from the public cloud providers. OpenStack helps Walmart with this.

For the layers higher in the software stack, like application servers, message systems, search indexes, caching and databases, the enterprise will likely choose an open source package, perhaps nginx, node.js, RabbitMQ, PostgreSQL, Elasticsearch, Kafka, Redis, etc. Where a commercial offering exists with a paid enterprise version that provides incrementally better functionality than the open source community version, the enterprise will likely license that for their private data centers. Examples of software services that have both a commercial and open source community option (incidentally usually maintained by the same company) are MongoDB, Confluent, Elastic, Redis Labs, etc.

For these components, the deployment in a private data center doesn’t represent much impact. The enterprise would still be inclined to license a commercial version of the open source software package and install it in their data center. They will also likely purchase some sort of support contract, so that they have access to the expertise of the commercial vendor and project maintainer. This is the same decision whether they use a public or private cloud. Additionally, the same open source packages generally offer a cloud-hosted version of their software on the major public cloud providers. Since the programming interface is the same, the enterprise doesn’t need to make any changes to their application in order to use either self-hosted or public cloud-hosted configurations.

For external services that the enterprise chooses not to duplicate themselves, the deployment in a hybrid model would have little to no impact. This applies to externally managed software services that would not have a self-hosted model and sit outside the data center by default. Examples of these types of services are observability (Datadog), CPaaS (Twilio), endpoint security (Crowdstrike), identity (Okta), CDN/DDOS (Cloudflare), ITSM (ServiceNow). Unless the enterprise is planning to shift this functionality to some in-house solution (whether through open source or their own custom software), these providers would experience minimal impact by a hybrid model. For Walmart’s use of stores as edge locations, that is somewhat unique to their use case. In a general sense, it promotes the advantages of deploying applications to the edge, benefiting edge compute providers.

Finally, Walmart recognizes the special case that large data processing and AI/ML represent. They are willing to outsource that function, acknowledging that it would be very costly and complex to duplicate in their own private data center.

All in all, the example from Walmart is instructive for investors because it provides a concrete example of a major enterprise’s cloud journey and where they expect to gain operational leverage. Aside from the low-level software infrastructure services (like compute, storage and network), an enterprise’s choice to run some or even most of their application workloads on their private data centers does not automatically result in less utilization of commercial software infrastructure packages and services. That utilization would likely remain comparable to a full deployment on the public clouds. Savings realized from running their own compute, storage and network hardware might even be redeployed into new high-value digital initiatives like new customer applications, more automation or better harnessing of big data through analytics and machine learning.

NOTE: This article does not represent investment advice and is solely the author’s opinion for managing his own investment portfolio. Readers are expected to perform their own due diligence before making investment decisions. Please see the Disclaimer for more detail.