Scale capacity and performance non-disruptively

Unlike competing designs that merely cascade additional SSDs in so-called JBOF (just a bunch of flash) enclosures, Gemini all-flash arrays scale performance and capacity linearly. Our patent-pending Distributed Cache Architecture adds cache in tandem with capacity, while Gemini’s patented internal fabric adds dedicated bandwidth for each new flash drive automatically.

From 6 TB to 48 TB

raw in just 2U

Expand on-demand

without any disruption

Manage exabytes of flash

from one interface

Dennis Martin
President, Demartek

Limitations of Traditional Architectures

Traditional storage designs rely on controller-resident NVRAM to accelerate writes, but this design mandates that controllers mirror the writes to safeguard data against controller failures. These designs also require batteries to protect the data if power is lost. These approaches introduce architectural complexity that add cost and risk to the storage system.

Distributed Cache Architecture

Nimbus Data’s approach, on the other hand, distributes the write cache across the flash drives within the system, relying on its patented non-blocking fabric to ensure low latency. This eliminates the need for controller-based NVRAM and controller-to-controller mirroring, reducing hardware costs and improving reliability. Additionally, since no data is acknowledged prior to being committed to persistent flash memory, no batteries are required, reducing maintenance costs and risk. Finally, adding flash drives automatically adds cache, helping ensure linear performance improvements as capacity is added.

On-demand Capacity and Performance

Gemini systems scale to 48 TB of raw flash in a 2U chassis. Accounting for data protection overhead and up to 10:1 savings with deduplication and compression, Gemini all-flash arrays support up to 400 TB of effective capacity in just one system, delivering 3-10x more capacity in a fraction of the space of competing systems. Additional flash drives can be added on-the-fly to expand existing volumes instantly and without downtime.