How Do Cloud‑Native Storage Solutions Stack Up? Performance Insights

1. Introduction

Cloud‑native has become the primary choice for application developers when designing architectures, allowing them to focus on business logic while reducing development burden.

The agility, scalability, reliability, and high availability of applications are now jointly handled by infrastructure software and operations teams, who must quickly respond to deployment, scaling, and migration requests while maintaining system stability.

2. Current Situation

Choosing the right infrastructure software and storage solution is a major challenge for operations leaders.

3. Storage Systems in Cloud‑Native Scenarios

Storage has always been a core component of infrastructure software. Over the past 30 years, deployment environments have evolved from standalone physical machines to virtualized clusters, large‑scale clouds, and today’s Kubernetes platforms. This evolution demands higher agility from storage systems, which must support rapid provisioning, scaling, and automated operations.

4. Cloud‑Native Storage Options

Local Disk : The simplest and historically common method, involving direct attachment of disks to servers, RAID configuration, and mounting to containers. It offers the shortest I/O path and high performance, but lacks node‑level high availability, limited scalability, and high operational overhead.

Centralized Storage : Provides remote shared storage, solving high‑availability issues and reducing space waste. However, its controller‑based architecture can become a performance bottleneck under heavy concurrency, and it was not designed for the agility required by cloud‑native workloads.

Distributed Storage : Designed to overcome the limitations of centralized storage, offering horizontal scalability, superior performance, and high availability. While it brings many architectural benefits, implementation complexity and production‑grade stability remain challenges.

5. Requirements of Stateful Cloud‑Native Applications

Stateful cloud‑native workloads are largely represented by cloud‑native databases. Two typical database types are considered:

Transactional (OLTP) Databases such as MySQL and PostgreSQL demand high reliability, low latency, and high throughput; they typically use block storage for optimal performance.

Analytical (OLAP) Databases focus on large data volumes and cost‑effective storage, often leveraging object storage (e.g., S3) with erasure coding.

6. New Challenges in Multi‑Cloud Environments

Enterprises increasingly adopt multi‑cloud strategies, using public clouds for less sensitive, high‑traffic workloads and private clouds for sensitive, high‑availability services. This diversity imposes heavy operational burdens on infrastructure teams.

Cloud‑native storage must run seamlessly across public and private clouds, offering containerized deployment, automated operations, and declarative interfaces to reduce operational complexity.

7. How to Choose Cloud‑Native Storage

Reliability, performance, and high availability are fundamental.

Minimize environment dependencies (kernel, hardware, network).

Avoid excessive resource consumption; storage should not dominate CPU or memory.

Support declarative, automated operations (online scaling, upgrades, fault recovery).

Integrate with the cloud‑native ecosystem (monitoring, alerting, logging).

8. Performance Comparison of Cloud‑Native Storage Systems

Four cloud‑native storage solutions—IOMesh, Longhorn, Portworx, and OpenEBS—were benchmarked using a MySQL instance driven by sysbench‑tpcc. The test environment consisted of three worker nodes, each equipped with two SATA SSDs, four SATA HDDs, and 10 GbE NICs.

The results show that IOMesh leads in both absolute performance and stability, making it suitable for core production workloads.

9. Conclusion

As the cloud‑native era advances, more systems adopt cloud‑native architectures, placing new demands on storage. Both databases and storage systems are evolving toward cloud‑native designs, and we can expect widespread adoption of cloud‑native databases and storage solutions in future data centers.