The CrowdStrike outage highlighted the vulnerabilities of digitalization across the global economy. For fleets, whose most important digital asset might be the transportation management system, the blackout underscores the importance of a resilient TMS deployment.

When ensuring a fleet’s TMS deployment is as secure as possible, one important choice is whether to host the software on-premise or in the cloud. On-premise might be the more expensive, customizable option—but it isn’t necessarily the best for insulating a business from outages like CrowdStrike’s.

A similar choice lies in choosing between single-tenant and multi-tenant architectures. Again, single-tenant is pricey and highly customizable—but that extra cost isn’t always paying for quicker disaster recovery.

What caused the IT outage?

On July 19, cybersecurity company CrowdStrike distributed an erroneous update to all Windows devices using its Falcon Sensor security software. The update brought critical problems to computers running the software.

Specifically, the update broke Windows computers’ startup processes. The updated process asked computers to access an invalid memory address in kernel mode, causing catastrophic failure.

Most software updates don’t break the software, let alone an entire computer. How did CrowdStrike manage to distribute such a harmful update? Simply put, one of the most important cybersecurity companies in the nation had shockingly weak quality assurance. User forums claim CrowdStrike had pushed similar erroneous updates affecting specific Linux operating systems in April, May, and June.

The solution was, at its simplest, to reboot the computer in safe mode, find the broken files from the update, and delete them. After deleting the appropriate files, the computer would boot successfully.

The update impacted roughly 8.5 million Windows computers, according to Microsoft. That’s less than 1% of all Windows machines globally. However, computers running Falcon Sensor included airlines, hospitals, and truck fleets, shocking the global economy.

Challenges for on-premise and single-tenant architectures

When compared to cloud-based software deployments, on-premise software might be attractive for its customizability—but it comes at the cost of disaster response agility.

“I think what is interesting to see is that the institutions that are affected by [the CrowdStrike outage] are airlines, governments, and big hospitals,” Hans Galland, CEO and founder of BeyondTrucks, told FleetOwner. “These are typically institutions that run their software in on-premise service.”

With on-premise solutions, the software is installed and run on a company’s own hardware infrastructure. This allows the company more direct control over the exact type of hardware, its location, its software, and its data. This is significantly more expensive than cloud-based deployments but comes with greater control and customizability.

Larger businesses that choose on-premise solutions tend to do so for the security, reliability, and efficiency of an independent solution. Third-party cloud hosts like Azure might inflict several service outages every month, bringing serious losses for the business.

“A lot of fleets, and usually the largest fleets, still have an on-premise system,” Galland said. “I would say 80% of fleets with 100 drivers or more either use private cloud or on-premise. It’s very significant.”

However, on-premise deployments may have a much harder time competing with the maintenance and disaster recovery efficiencies offered by cloud deployments. On-premise deployments might only have a single IT person manually updating each machine. Cloud providers, able to afford larger dedicated maintenance staff, generally have greater agility in responding to quick patches or updates. When a cybersecurity company suddenly breaks 8.5 million Windows computers, greater agility can make a significant difference.

The CrowdStrike outage shows similar tradeoffs for fleets choosing between single-tenant and multi-tenant software deployments. Single-tenant architecture hosts a software application and supporting infrastructure for a single customer. Multi-tenant architecture, by contrast, hosts multiple customers within a single software instance. Single-tenant deployments might be more customizable (and more expensive), but they are also likely to face slower patch rollouts.