Setting the Stage

In today’s cloud environments, storage is more than just infrastructure—it’s the heartbeat of mission-critical applications. From seismic interpretation in the energy sector to other high-performance scientific workloads, organizations need to decrease time-to-decision, whether in identifying economically recoverable energy resources or accelerating other data-intensive scientific discoveries.

That’s why Daniel Gomes, a Principal Solution Engineer at Qumulo specializing in Oil & Gas workloads, constructed a head-to-head performance benchmark between Cloud Native Qumulo and Amazon FSx for NetApp ONTAP (FSxN). Daniel designed these tests to reflect real-world conditions, using client patterns and dataset sizes commonly seen in oil and gas interpretation workloads.

The results: Qumulo consistently outperformed FSxN across single-client and multi-client scenarios, especially as concurrency and read intensity increased. The methodology was designed to capture conditions characteristic of oil and gas interpretation applications – including, but not limited to, those used by one of the largest global providers of geoscience and seismic interpretation platforms – as well as other industries with similar file system performance demands. Daniels’s goal was to create reliable and repeatable benchmarks that can inform both customer expectations and technical decisions.

Why Baseline Testing Matters

Baseline testing is essential because the risks of unknown behaviors, unpredictability, and extended project timelines directly impact the time-to-decision across many project objectives. Baselining sets realistic expectations for performance, helps teams pinpoint bottlenecks, and enables faster troubleshooting and resolution. By benchmarking Qumulo’s capabilities against NetApp’s in AWS, this benchmark delivers a transparent, practical comparison between the two leading providers of unstructured data services in the subsurface interpretation ecosystem. This benchmark delivers exactly that—a transparent, cloud-native baseline comparison of Qumulo and FSxN on AWS.

Testing Design

• Clients: Red Hat Enterprise Linux (RHEL) on AWS EC2 g4dn.16xlarge instances
• Tooling: fio (v3.36) with industry-accepted I/O profiles
• Datasets: 50GB per test, reflecting typical energy sector project sizes (10GB–150GB)
• Protocols: NFSv3 with standardized mount options across both platforms
• Workloads: Sequential and random reads/writes under single-client and six-client concurrency

On the storage side:

• Qumulo: A three-node cluster with NeuralCache, NVMe and gp3 caching, backed by S3 Intelligent Tiering.
• FSxN: Dual-controller HA system tested at 1.5, 3, and 6 GBps throughput tiers with maximum cache enabled.
• 50% Less Cost: The Qumulo system used for testing would retail for ~50% less than the price of the FSxN solution.

Results

Key Findings

1. Qumulo Scales With Clients
As client load increased, Qumulo’s performance grew, while FSxN plateaued.

• Six-client sequential read (64k):
  • Qumulo: 11.57 GB/s (190K IOPS)
  • FSxN (6 GBps tier): 4.98 GB/s (82K IOPS)

2. NeuralCache Delivers Consistent Advantage
Even when FSxN was provisioned with maximum cache, Qumulo’s NeuralCache architecture outperformed in read-heavy workloads and maintained concurrency without complexity.

3. Writes: Close, But Qumulo Wins at Scale
FSxN showed strength in some single-client writes, but Qumulo pulled ahead as concurrency increased:

• Six-client random write (64k):
  • Qumulo: 1.65 GB/s (28K IOPS)
  • FSxN (6 GBps tier): 1.19 GB/s (20K IOPS)

4. Simplicity and Speed

• Deployment: Qumulo stood up in ~5 minutes vs. FSxN’s ~19 minutes
• Scaling: Adding Qumulo nodes or adjusting EC2 types takes minutes, compared to an hour or more for FSxN throughput changes
• Manageability: Qumulo scales seamlessly in a single namespace; FSxN requires multiple file systems and Storage Virtual Machines to match performance

What This Means for Enterprises

For organizations with demanding, multi-user workloads like oil and gas interpretation applications, Qumulo offers:

• Higher throughput and IOPS under real-world concurrency
• Lower operational complexity with faster deployment and scaling
• Future-proof simplicity for teams managing petabyte-scale growth
• Lower cost for teams looking to maximize their budgets

In contrast, FSxN requires more manual tuning and still lags behind Qumulo in critical performance areas.

Final Word

This benchmark on AWS clearly shows: Qumulo is the best choice for scalable, high-performance file data workloads in the cloud for seismic interpretation.
Whether you’re in the energy sector doing seismic interpretation or any industry where concurrency and read performance are crucial, Qumulo provides the throughput, agility, and simplicity needed to keep your applications running at peak levels. Qumulo has already enabled petabyte-scale migrations from NetApp cloud solutions into the Cloud Native Qumulo (CNQ) architecture. After making this switch, organizations using CNQ report lower operating costs, faster decision-making, and greater technology flexibility with new applications and AI tooling.

Want to see the full benchmark results and methodology? Contact us for details at info@qumulo.com.