Why CrowdStrike was a wake-up call for businesses’ edge computing strategies
By Said Ouissal, CEO and co-founder of Zededa
The recent CrowdStrike incident was a stark reminder of the challenges enterprises face in deploying and managing software at scale. When a faulty software update exposes the fragility of our interconnected global web of software, IT systems and devices, companies are forced to take a hard look at their traditional software processes.
Perhaps the biggest overall lesson from the outage is that decentralizing IT needs to be a top priority for organizations managing critical or industrial infrastructure. The growth of edge computing and the increasing reliance on AI and machine learning at the edge means that building enterprise technical architectures to meet the resiliency requirements of distributed environments is more important than ever.
While no system or process is perfect, the details from the CrowdStrike coverage starkly illustrate where traditional efforts failed and what experts managing edge computing deployments can learn from those mistakes.
1. An over-dependence on the underlying operating system
Like many endpoint solutions, the CrowdStrike agent relied heavily on the underlying operating system for recovery and resiliency. This approach is unsustainable in edge computing, where devices may be in remote or hard-to-reach locations. Edge solutions must be isolated from the underlying OS and from each other to improve resiliency and limit the impact of any single failed update.
2. A lack of robust rollback mechanisms
The inability to easily roll back to a prior operating system version left many systems bricked and unreachable. For edge devices, the underlying OS must be able to test any update and automatically revert to a previous version if something goes wrong.
3. Limited remote recovery options
The only way to fix the CrowdStrike error was to manually reset each IT endpoint, which cost airlines, healthcare organizations and others way too much time. Manual intervention to reset affected endpoints is simply not feasible at the edge. With devices potentially scattered across vast geographical areas or in hazardous locations, all operations must be accessible remotely, regardless of the situation or condition.
These challenges underscore why edge computing requires a fundamentally different approach. Building a robust edge computing platform means more than just creating a management agent. It requires rethinking the entire stack from the ground up. EVE-OS, an open-source operating system within The Linux Foundation’s LF Edge organization, was developed specifically for the unique challenges of distributed environments. EVE-OS incorporates features like:
– Autonomous rollback capabilities that detect and revert to the last known working version without manual intervention
– Isolation of applications and services from the underlying OS, as well as from each other, to enhance resilience
– Comprehensive remote management capabilities for all aspects of the system
– Automatic verification of all edge elements using measured boot and remote attestation to detect changes
As the scale of edge deployments grows globally, organizations must leverage technology such as EVE-OS to ensure that their critical infrastructure can withstand the inevitable challenges that will arise. The future of edge computing depends on our ability to deploy and manage software at scale with minimal risk of widespread outages. This requires a shift in thinking from traditional centralized IT models to more robust, self-healing systems designed for the unique challenges of the edge.
As we move forward, let’s use this incident as a catalyst for change. By embracing new approaches and technologies specifically designed for distributed environments, we can build a more resilient and reliable edge computing ecosystem for all.
About the author
Said Ouissal is CEO and founder of Zededa. Zededa offers an enterprise edge cloud-based service that delivers visibility, control, and protection for distributed edge gateways, applications, and networks.
Evolutionary fault tolerance
Comments