Rob Pankow 
Introduction
At simplyblock.io we have conducted a series of performance tests on our NVMe-oF storage solution on AWS. Simplyblock’s cloud-native architecture delivers high throughput, low latency, and linear scalability for organizations and service providers running data-intensive workloads. This article covers our benchmark results and what they mean for real-world deployments.
Performance at Scale
Simplyblock’s NVMe-oF storage solution is designed for high-throughput, low-latency workloads such as machine learning, analytics, genomics, and databases. In production deployments on AWS, simplyblock has demonstrated:
- >128 Gbit/s total throughput in the MLPerf® Storage v2.0 benchmark from a minimal 3-node cluster on standard cloud infrastructure
- 3 GiB/s per GPU — industry-leading throughput per accelerator for AI/ML training workloads
- 200% performance improvement in the ClickBench analytical benchmark compared to standard EBS-backed deployments
- Sub-100 microsecond average access latency, enabling applications to process data with minimal delay
These results are achieved by harnessing NVMe-oF technology across standard EC2 infrastructure — no specialized hardware required.
Linear Scalability and Flexibility
One key feature of simplyblock’s solution is its linear scalability. Adding nodes proportionally increases both storage performance and capacity, allowing organizations to scale to match growing workloads without bottlenecks or re-architecting. This has been validated in the MLPerf benchmark, where throughput scaled efficiently across accelerator counts.
Low Latency and Rapid Data Access
Simplyblock delivers low-latency data access over standard Ethernet using NVMe/TCP. Critical applications — from PostgreSQL to real-time analytics — benefit from consistent, predictable I/O performance. Combined with multi-layer caching and intelligent tiering, simplyblock keeps hot data close to compute while automatically offloading cold data to lower-cost storage.
Data Protection and Reliability
Simplyblock prioritizes data protection through built-in redundancy, erasure coding, and RAID-like protection. Data remains accessible even in the event of node or drive failures, ensuring high availability for sensitive and mission-critical workloads.
Conclusion
Simplyblock’s AWS deployments have demonstrated industry-leading results across multiple independent benchmarks — including MLPerf Storage and ClickBench — validating its position as a high-performance, cloud-native alternative to Amazon EBS for demanding workloads. With linear scalability, sub-millisecond latency, and >128 Gbit/s throughput at scale, organizations can confidently run data-intensive applications on AWS without compromising performance or paying for overprovisioned storage.
Simplyblock serves as an alternative to Amazon EBS storage, fully compatible with the AWS platform and deployable natively on EC2 instances. It also runs in any other cloud or on-premises environment, including OpenStack, OpenNebula, Proxmox, or VMware — as an alternative to Ceph or vSAN.
Questions and Answers
How does Simplyblock perform compared to AWS EBS?
In benchmark tests, simplyblock’s NVMe-over-TCP volumes significantly outperform AWS EBS gp3 and io2 in both IOPS and latency. When deployed on EC2, simplyblock provides higher performance with lower overhead. Check our AWS EBS pricing guide for context on potential cost savings.
Can Simplyblock reduce storage costs on AWS?
Yes. Simplyblock enables EBS volume consolidation and replaces expensive io2 volumes with high-performance NVMe-backed alternatives. Combined with flat pricing and better efficiency, this results in lower overall storage costs.
Is Simplyblock compatible with Kubernetes on AWS?
Absolutely. Simplyblock integrates with Kubernetes using a CSI driver, allowing seamless provisioning of persistent volumes with high IOPS, ideal for stateful workloads like databases and streaming platforms on EKS.
What use cases benefit most from Simplyblock on AWS?
High-performance applications such as PostgreSQL, MongoDB, and analytics engines benefit most. simplyblock for PostgreSQL and MongoDB delivers low-latency storage ideal for transactional and real-time workloads.
How does Simplyblock use NVMe over TCP on AWS?
Simplyblock runs NVMe-over-TCP over standard EC2 infrastructure, enabling fast, scalable block storage without the need for special hardware. Learn more about how NVMe over TCP works and why it’s ideal for cloud environments.
