Walmart scales stupendous global network on open source

  • Open source helps the global retailer build a performant network spanning 10,000-plus locations

  • Walmart emphasizes cost-efficiency and network visibility, leveraging open-source solutions like SONiC

  • The retailer is exploring future technologies including private 5G and Wi-Fi 8

From a single Arkansas store in 1950, Walmart has grown to operate one of the world's biggest and most sophisticated networks — leveraging open source to run its global business.

Dave Temkin, Walmart VP infrastructure, laid out the scope of the network and business it drives, by the numbers, in a keynote at the Linux Foundation Open Networking & Edge (ONE) Summit in San Jose, Calif., Tuesday: 

  • More than 10,000 stores and e-commerce websites
  • In 19 countries
  • 1,000 global supply chain locations
  • 240 million customers weekly
  • The retailer employs 2.1 million associates, each with a mobile device
  • The network runs 100,000-plus routers and switches
  • 75% of the switches utilize power-over-Ethernet
  • The network also runs 450,000+ wireless access points
  • Network devices consume more than 4.5MW PoE in Walmart U.S. per week

Also, Walmart sells 1.5 billion pounds of bananas per year (“random stat,” Temkin commented).

Temkin joined Walmart four months ago, as VP of infrastructure. Previously, Temkin headed the content distribution network (CDN) for Netflix (“everything that happens after you press play,” he explained) and headed infrastructure and cloud for Imperva, a cybersecurity provider acquired by Thales for $3.6 billion.

“People say ‘Dave, you worked at those big networks, why would you go into retail?’” Temkin said. “The answer is scale.”

Black-and-white photo of a small shop on a midcentury American street. The shop sign above the windows, black letters on a white background, says "Walton's 5-10." Fantastic midcentury sedans are parked front-end-in on the street in front.
Walton’s 5 & 10 in the early 1950s in Bentonville, Ark. Walmart founder Sam Walton opened the store May 9 ,1950.  (Walmart)

The Walmart network is indeed enormous, said Arpit Joshipura, the conference host and SVP/GM Networking + Edge + IoT for the Linux Foundation, during a conversation with Fierce Network during a break in the conference.

“By the numbers, it’s one of the largest enterprises and one of the largest service provider networks, with so many deployments and nodes,” Joshipura said. “I was surprised at the scale at which they run. Each of their buildings is an edge node. It’s incredibly complex, running such a large network in real-time."

Innovating on that network is challenging, Temkin said. “We have to keep this business operating 24x7,” he said. “We have an infrastructure that has grown organically over the years. We can’t just come in and say, ok, we’re going to rip something out and move to something new.”

Like painting a bridge

Visibility across the entire chain is a challenge, according to Temkin.

“I have to know what’s going on across the chain, its 10,000-plus locations and e-commerce platforms, at all times,” Temkin said. Walmart also needs to reduce operational costs, and manage multiple vendors and platforms, maintaining feature parity, bug mitigation and security because “there is nothing off-the-shelf that solves the problem,” he said.

With 500,000 network elements, lifecycle management is a constant job. “I’ve started thinking about this as being like painting a bridge,” Temkin said. “By the time I’m done modernizing that entire chain, I’m going to have to come around and start all over again, because there is no way for me to roll things out in parallel in that large an environment.”

Key variables include achieving consistent visibility across platforms, reducing the operational overhead of managing network operating system (NOS) versions, traffic management, scalable load balancing, device lifecycle management at Walmart scale, feature management and efficiency, maintaining vendor agnosticism to the extent that’s practical and cost management.

“At Walmart, we talk about EDLC — everyday low cost. Every dollar that I save in the network is going out to saving our customers money, and that’s paramount to everything else,” Temkin said.

Open source is key to maintaining optimal operations at low cost. Walmart has deployed the SONiC open source network operating system in its data centers and would like to be able to expand that deployment to its stores, distribution centers and campuses. SONiC has “yielded incredible results in operational excellence, reliability and engineering cognitive load,” Temkins said. The retailer plans to “grow and hire talent for SONiC development and community contribution.”

Walmart is also heavily involved in L3AF, an open source orchestration platform and lifecycle manager for eBPF programs. At Walmart, L3AF and eBPF are implemented for a range of network functions, including observability, network functions (such as prioritizing payment in stores) and security.

The retail giant sees generative AI (GenAI) as a tool to analyze real-time network events. Its 500,000-plus network devices generate billions of log lines daily. GenAI can help predictably understand network behavior, scalability and avoid outages, Temkin said during a conversation with Fierce after his presentation Tuesday. 

Warehouse robots and drone delivery

Walmart’s innovation isn’t limited to the data center. The company is involved in high-profile retail projects such as store robots and drone delivery. These initiatives require cloud power and strong, resilient networks, Temkin said.

For communications, Walmart is looking into private 5G in the short term and the emerging Wi-Fi 8 standard in the long term. The company sees private 5G as an interim measure because 5G is optimized for carrier needs, while Wi-Fi is more scalable and open.

“To me, private 5G is kind of a bridge technology to get us to a more cohesive strategy where we can have one technology that runs communications,” Temkins said.

For compute, Walmart operates what it calls a “triple cloud network.” Walmart runs workloads on Microsoft Azure, Google Cloud Platform and its own private cloud. The company selects which workloads to run on which cloud based primarily on availability and locality, to maximize performance.

Walmart’s infrastructure is impressive both in its scale and commitment to open source, said Roy Rubenstein, an analyst with LightCounting. “They can’t afford to just go with one vendor and buy what they offer. So they’re really interested in open-source software,” Rubenstein said. “They work with vendors, but they want the vendors to be open source, use open interfaces and work with Walmart on what they want to do. And they’ve got the capability to run with open source, because they’re large.” 

Roy Chua, founder and principal analyst at AdvidThink, agreed. A large organization like Walmart (or Capital One) can pioneer open source and innovate on networking and the cloud because it has sufficient internal capabilities, expertise and ability to recruit and retain networking and open source experts, he said. Such a large organization can also demonstrate ROI for their investment because of their scale.

“Other businesses tend to watch and then mimic later,” Chua said, “Either finding a vendor or systems integrator who can help them implement a blueprint using open source, or benefiting from a commercial vendor that codifies these learnings and best practices into product.”

But even an organization with the scale of a Walmart — or Amazon, Google or another hyperscaler — can’t develop all the technology it needs internally, as Rubenstein noted. Open source makes it possible for companies to collaborate on their IT needs.

Don't miss our upcoming Cloud-native 5G virtual event, which will feature speakers from AWS, Microsoft, and Google Cloud. Register for your free pass here.