Validated testing shows resilient AI data delivery with Dell ObjectScale and F5 BIG-IP

Paul PindellPrincipal Solutions Architect | F5

Mark MengerSolutions Architect | F5

Hunter SmitSenior Manager, Product Marketing | F5

AI infrastructure does not slow down only when compute is constrained. It slows down when data cannot move fast enough, consistently enough, or reliably enough to keep AI workflows running. For model training, fine-tuning, and retrieval-augmented generation (RAG), S3-compatible object storage has become a critical foundation. Massive datasets land in object stores. Training jobs read them at scale. RAG pipelines depend on fast, predictable access to continuously changing data.

That makes the reliability, scalability, and security of the data path between storage and compute a production concern.

AI needs a stronger data path

Dell ObjectScale is designed for enterprise object storage at AI scale, helping organizations manage large-scale data collection, generative AI training, and modern unstructured data workloads. But as AI environments move from pilot projects into production, storage performance alone is but one piece of the whole architecture. Enterprises also need a delivery layer that can optimize traffic, maintain availability, enforce policy, and keep pipelines moving when the environment is under stress.

That is the role of F5 BIG-IP in AI data delivery.

F5 BIG-IP provides a programmable, high-performance control point for S3-compatible object storage. It sits between S3 clients and storage environments such as Dell ObjectScale to intelligently manage S3 traffic, abstract backend storage endpoints, monitor health, and steer data flows based on availability, performance, and policy.

For architects, that means applications and AI workflows can remain decoupled from the underlying storage implementation. For executives, it means AI infrastructure can be operated with greater reliability, control, and resilience.

SecureIQLab validates the architecture and results

F5 recently completed engineering lab testing with Dell ObjectScale storage, using F5 VELOS and BIG-IP with a Dell ObjectScale S3 cluster. The results were then validated by SecureIQLab and are being released in a new, independent report.

The testing focused on two practical questions enterprise architects will ask before putting a delivery layer in front of AI storage.

1. Can BIG-IP be inserted into the S3 data path without compromising throughput, and possibly improve throughput metrics for less than perfect network conditions?
2. Can BIG-IP add resiliency when the storage environment is under adverse conditions?

The answer to both was yes.

BIG-IP preserved performance

In throughput testing, F5 validated that BIG-IP can be inserted in front of Dell ObjectScale with a “do no harm” performance baseline when correctly tuned. Using a refined custom TCP profile, BIG-IP achieved near-parity with the no-BIG-IP baseline in most tested conditions.

That matters because many storage buyers assume an application delivery controller will add consequential overhead to a high-performance data path. These results show that BIG-IP can be introduced as a high-value control point without becoming the bottleneck.

BIG-IP improved latency-sensitive throughput

In certain high latency scenarios, BIG-IP did more than preserve performance. It improved it.

With 8 KiB objects at 10 ms latency, the custom TCP profile delivered 19.61 Gbps versus 3.97 Gbps without BIG-IP, a 393.9% improvement. With 100 MiB objects at 75 ms latency, it delivered 588.49 Gbps versus 316.96 Gbps without BIG-IP, an 85.7% improvement.

These results reflect the realities of distributed AI infrastructure—where latency, object size, connection behavior, and TCP efficiency can materially affect throughput.

The takeaway is not that every environment will see the same gains. The takeaway is more important: when AI data pipelines are adversely affected by latency, connection churn, or TCP windowing inefficiencies, BIG-IP optimizes the data path between clients and Dell ObjectScale.

Resiliency is the larger value

Performance is only the first part of the story.

The SecureIQLab-validated testing also examined resiliency. During a SYN-flood scenario with live S3 traffic, the BIG-IP virtual server continued processing high-volume S3 traffic without errors. During a node outage test, after 100G ports on an ObjectScale node were shut down, BIG-IP continued processing S3 traffic without errors while the failed node was confirmed unreachable.

For AI infrastructure teams, that is the larger architectural point. In production, the data path has to survive real operating conditions. Nodes fail. Networks degrade. Traffic surges. Attack traffic competes with legitimate traffic. Maintenance windows collide with business demand.

Without a dedicated control point, these events can turn into stalled training runs, failed RAG queries, idle GPUs, and unpredictable user experiences.

Production AI requires control points

BIG-IP helps to address those risks by creating a resilient delivery tier for AI data. It can monitor storage health, route around unavailable nodes, preserve service continuity, and apply traffic controls before the storage cluster becomes unstable.

It also provides a place to consistently enforce security and governance, including TLS handling, DDoS mitigation, S3-aware policy enforcement, and future data-plane controls that help clients obtain predictable backpressure instead of random timeouts.

For enterprises deploying AI storage clusters, this is a practical architecture pattern: pair scalable, S3-compatible object storage with an intelligent delivery layer purpose-built for performance, policy, and resiliency.

The takeaway for AI leaders

For AI leaders, the business value is straightforward. Faster and more reliable data movement helps keep expensive infrastructure productive. Resilient storage delivery reduces operational risk. A programmable control point gives teams more confidence as AI workloads scale from experimentation to production.

AI success depends on more than the model. It depends on the system around the model.

Download the SecureIQLab report to see the validated findings for F5 BIG-IP with Dell ObjectScale and learn how F5 helps make AI data pipelines faster, more resilient, and ready for production.

Also, be sure to watch our webinar, “Enhancing AI Performance with Dell Infrastructure.”

About the Authors

Paul PindellPrincipal Solutions Architect | F5

Paul Pindell is a Principal Architect at F5, where he works in Technology Alliances and oversees technical partnerships across F5’s product portfolio. Since 2019, he has also played a key role in corporate strategy initiatives, including helping launch F5’s first innovation project. Paul led F5’s AI strategy tiger team, focused on defining how F5’s existing products can help secure and deliver AI applications. He later helped develop F5’s AI Reference Architecture, outlining the core building blocks, challenges, risks, and partner solutions required to support AI application delivery and security. Paul is a frequent keynote and breakout speaker at industry events, including OPI Summit, OCP Global Summit, Intel InnovatiON, Open Source Summit, Red Hat Tech Exchange, and multiple VMworld conferences across the U.S. and EMEA. He is also the founder and maintainer of open source projects, serves as Chair of the Outreach Committee for the Open Programmable Infrastructure Project, and previously led data center operations for Symantec’s Antivirus Response Data Centers and security software testing labs.

More blogs by Paul Pindell

Mark MengerSolutions Architect | F5

Mark Menger is a Solutions Architect at F5, specializing in AI and security technology partnerships. He leads the development of F5’s AI Reference Architecture, advancing secure, scalable AI solutions. With experience as a Global Solutions Architect and Solutions Engineer, Mark contributed to F5’s Secure Cloud Architecture and co-developed its Distributed Four-Tiered Architecture. Co-author of Solving IT Complexity, he brings expertise in addressing IT challenges. Previously, he held roles as an application developer and enterprise architect, focusing on modern applications, automation, and accelerating value from AI investments.

More blogs by Mark Menger

Hunter SmitSenior Manager, Product Marketing | F5

Hunter Smit is a senior manager of product marketing for solutions at F5. He leads solution-level go-to-market efforts focused on AI use cases and expansion into new markets, including initiatives for AI data delivery. Hunter holds undergraduate and graduate degrees in business administration from Whitworth University and is a member of the Whitworth School of Business Advisory Board.

More blogs by Hunter Smit