FlashBlade Surpasses 3 Trillion Objects, Reimagining What’s Possible for Scale

On this day in 1947, US Air Force test pilot Chuck Yeager broke the sound barrier in a Bell X-1, at the same time shattering expectations that doing so might shred both the plane and himself to pieces. 

Instead, the moment was almost… anticlimactic. The Bell X-1 had been engineered so perfectly for top speed that hitting Mach 1 felt “smooth as a baby’s bottom.” The technology made it feel effortless.

The anniversary of this achievement is the perfect moment to share a similar landmark achievement that quietly occurred in a Pure Storage customer’s data center last month: 

A single Pure Storage® FlashBlade//E™ array stored over 3 trillion objects, and that number is still climbing. 

It’s all thanks to a client’s challenge and an engineer’s ingenuity, leading to a simple test of the limits of FlashBlade that far exceeded its goal. 

Why This Epic Breakthrough Matters Now More than Ever

Just as Yeager’s flight wasn’t solely about speed but paving the way for future innovation, this feat is not just about reaching 3 trillion objects. The real breakthrough is in how this performance translates into outcomes for customers.

“Customers want to see real numbers in the real world and not just read data sheets and trust the claims,” explains Russell Pope, Consulting Field Solutions Architect on the project. “The original challenge was to prove a FlashBlade could manage 500 billion objects, and when I put together a series of tools to prove our claims, I wasn’t disappointed in the results… especially when we delivered six times the original task.” 

The Bell X-1 solved problems beyond breaking the barrier, like turbulence and control, leading to innovations in aerodynamics, materials, and practices that benefited all of aviation, not just ultra-fast jets. FlashBlade similarly rose to the occasion. In fact, the speed at which it created objects on day one is the same today—and holding. And, that number is continuing to grow with no degradation of performance because the test is still running.

Pope helped put that into perspective: “There is no degradation of performance. We’re creating object #3,873,348,996,481 just as quickly as we created object #1.”

Here’s where AI and the growing challenges of metadata performance and management will put storage infrastructures through a similar, monumental test. 

Object storage is critical to AI. It gives organizations the freedom to store, manage, and analyze unprecedented volumes of unstructured data beyond the limits of traditional file shares—goals that have proved out of reach on traditional infrastructure. AI is going to test those limits in pretty short order, but we’re confident this next challenge can be just as effortless on our platform. 

FlashBlade + AI: A ‘Mach 1 Moment’ for Pure Storage

Object storage systems today are housing billions or even trillions of individual objects. Each object that’s stored, regardless of its size, requires associated metadata to describe it: its name, creation date, size, access permissions, etc. Without a robust and highly scalable metadata engine, the ability to efficiently locate, manage, and retrieve these objects would quickly break down. The system would be unusable or force workarounds that would inhibit manageable scale. 

This makes metadata performance the critical measurement index that unlocks the massive data lake within the objects.

This challenged us to design a system that was not just highly scalable but also highly performant in both the metadata and data planes. With most legacy storage solutions, the metadata database is often a derivative of a general-purpose database, like Cassandra. We took a different approach with the design of FlashBlade, engineering our own purpose-built metadata subsystem. This test has shown that there is effectively no upper limit on the number of objects that can be stored. Granted, we’ll run out of physical capacity to store metadata, but that’s a limitation of the physical hardware and not the underlying software architecture.

Like the Bell X-1, FlashBlade//E was engineered to solve the “control and turbulence” for customers, empowering them to pursue ambitious outcomes with confidence… even if they haven’t reached that inflection point yet. The test demonstrated exactly how the FlashBlade architecture can scale while remaining reliable and performant as demands skyrocket. 

And we will only keep going. The sky’s no longer the limit—for AI and whatever comes next.

Stay tuned as we share the engineer’s side of the story next—with an in-depth look at the test methodology, specs, and FlashBlade under the hood.

Try FlashBlade

No hardware, no setup, no cost—no problem. Experience the self-service capabilities of FlashBlade.

Take a Test Drive