We are pleased to announce the Tech Preview of Lightbits RAM Reduction, an improved storage engine for Lightbits block storage designed to deliver far more SSD capacity per server while using a fraction of the memory. It is available today, on request, as an evaluation build, and it is the first step toward a production release.
Memory has become one of the hardest lines in an infrastructure budget. Demand from AI infrastructure has pulled DRAM supply away from general-purpose servers, prices have climbed steeply, and lead times have become difficult to plan around. Many of you have told us the same thing in different words: SSDs and CPUs bring challenges of their own, but nothing in the server quote has become as hard to justify as the RAM. For a storage platform, this matters twice, because memory also limits how much capacity a single server can hold.
That is true for Lightbits today. Our current engine keeps a detailed map of the data in memory, which is a large part of what makes it fast, and a high-density Lightbits server can be configured with up to 1 TB of RAM. At today’s prices, that is not the trade we want to ask our customers to make, so we set out to change it.
What We Are Building
The goal. RAM Reduction breaks the link between a server’s capacity and the memory it needs. We’re engineering toward up to 1 PB of raw capacity in a single server with 256 GB of RAM. Denser servers mean fewer servers for the same capacity, and with them less rack space, less power and less hardware to buy and maintain.
What stays the same. From the outside, this is the same Lightbits block storage: the same NVMe® over TCP data path, volumes, snapshots and clones, replication across servers, and APIs and tools. Your clients, your automation and your Kubernetes, OpenStack and virtualization integrations should not need to change.
A Tech Preview, for Evaluation Only
We want to be direct about what this build is. The RAM Reduction Tech Preview is for evaluation in labs and staging environments. It is not production-ready, not for production data, and not upgradable: plan it as a new, separate cluster you are prepared to reinstall.
It is also an early build. It installs on physical servers running RHEL 9, AlmaLinux 9, or Rocky Linux 9; its performance is not tuned yet, and a few data services arrive only with general availability. Every customer who receives the build also receives the full, current list of known limitations.
Evaluate It With Us
The Tech Preview is the first step, not the finish line: what we learn from evaluations feeds directly into the production release, and we will keep sharing our progress along the way.
The build is not a public download. Whether you run Lightbits today or are looking at it for the first time, if memory cost or server density is a constraint in your environment, reach out to us to request access. We will also sit down with you before you install it to review your hardware, plan an evaluation that reflects your workloads, and work through what RAM Reduction could mean for the size and cost of your next cluster.
Join us on the Lightbits User Community Hub to share your experiences and connect with our team.