AWS Gives Every Container Its Own Tiny Server
TL;DR: AWS has launched Nitro Sandboxes, a free feature that isolates container workloads inside their own lightweight virtual machines. This provides strong, kernel-level security for EKS and ECS with just a single configuration change.
Key facts
- Category
- Infrastructure
- Impact
- High
- Published
- Source
- InfoQ
Full summary
AWS now lets you run each container in its own secure micro-virtual machine for free, enabled by a single line of code.
Amazon Web Services has introduced a significant security upgrade for its container services called AWS Nitro Sandboxes. According to a report from InfoQ, the new feature allows developers to run individual container workloads—such as pods in the Elastic Kubernetes Service (EKS) or tasks in the Elastic Container Service (ECS)—inside their own dedicated micro-virtual machines (microVMs). This provides a powerful layer of isolation that was previously difficult and costly to achieve. The most compelling aspects of the launch are its simplicity and cost: enabling this enhanced security requires changing just a single configuration field, and AWS is offering the feature at no extra charge. This makes a high-security architecture accessible to any team running containers on AWS, from small startups to large enterprises, fundamentally changing the default security posture available on the platform.
At its core, Nitro Sandboxes leverages Firecracker, the same open-source virtualization technology that powers other AWS services like Lambda and Fargate. Traditionally, all containers running on a single host machine share that machine's operating system kernel. This shared-kernel model is efficient but creates a potential security risk; a vulnerability in one container could theoretically be exploited to “escape” and gain access to the host or other containers. Nitro Sandboxes eliminate this risk by wrapping each opted-in workload in its own lightweight Firecracker microVM. Each microVM has its own separate, minimal kernel, creating a hardware-enforced boundary between workloads. This means a compromise in one container is contained within its own sandbox, unable to affect its neighbors or the underlying host. For highly sensitive applications, AWS also offers a confidential mode that uses the AWS Nitro System to ensure code and data are protected even from the cloud provider itself.
This move by AWS is part of a broader industry trend toward stronger workload isolation in the cloud. The shared-kernel architecture of containers has long been a point of concern for security teams, especially in multi-tenant environments where applications from different customers run on the same infrastructure. In response, the open-source community developed projects like Google’s gVisor, which intercepts and filters system calls, and Kata Containers, which also uses lightweight VMs to isolate container runtimes. While effective, these solutions often required significant expertise to implement and manage. By integrating Firecracker-based sandboxing directly into its flagship container services and making it a simple, free option, AWS is democratizing this advanced security pattern. It lowers the barrier to entry for building secure, multi-tenant applications and puts pressure on other cloud providers to offer similarly integrated and easy-to-use isolation technologies.
For engineering and security teams, the decision to adopt Nitro Sandboxes involves weighing the powerful security benefits against a few documented trade-offs. The current implementation does not support GPU passthrough, which makes it unsuitable for machine learning or other GPU-accelerated workloads. There are also performance considerations: AWS limits the number of sandboxes to five per virtual CPU and notes that compute performance may be halved on instances with Simultaneous Multithreading (SMT) enabled. These limitations make it ideal for workloads that prioritize security over raw performance, such as running untrusted third-party code, processing sensitive data, or hosting SaaS applications. Teams should evaluate their specific use cases to determine if the performance impact is acceptable. Going forward, the industry will be watching to see how AWS evolves the service, potentially addressing current limitations and expanding its capabilities to cover an even broader range of workloads.
Why it matters
This gives developers a simple, no-cost way to achieve strong workload isolation, a major step up from traditional container security. It moves the security boundary from a shared kernel to a hardware-virtualized microVM, drastically reducing the attack surface for container escape vulnerabilities.
Business impact
Companies can now significantly harden the security of multi-tenant applications on AWS without added cost or complex re-architecture. This reduces the business risk of breaches from container vulnerabilities and simplifies compliance by providing stronger isolation guarantees between applications.
Tags
Related on Notifire
Related stories
Primary source: InfoQ
