Behind ATTO SiliconDisk: Solving Storage Challenges in a Brand New Way

This article is part four of a multi-part series that describes the impetus and technology behind ATTO SiliconDisk RAM-based storage appliance. For more details, visit

With the continued growth of artificial intelligence, data analytics and real-time data, enterprises face a crossroads. They must decide whether to continue with a traditional server architecture or adopt new solutions, likely unimagined until recently, to address the stresses of modern high-performance workloads.

ATTO SiliconDisk modernizes existing data center architectures where the gap between system RAM and storage is wide and storage is the bottleneck.For companies that rely on in-memory processing of data, SiliconDisk opens the door to accessing much higher quantities of mission-critical and hot data for RTD, AI, analytics, high-performance computing (HPC), big data operations and the like.

This capability is essential for real-time intelligence like fraud detection, operational intelligence and risk management, and data security monitoring. In the financial sector, there has been a push to standardize legacy data and integrate that into a modern analytics system. New data is collected every moment of the business day and is added to the store of existing data for processing and analysis. SiliconDisk could boost collection, merging and processing operations by being the first tier of storage beyond the servers where an I/O bottleneck would otherwise exist.

Architectures designed for artificial intelligence would benefit much the same way. Training and learning times could be improved if data needed for the models were available for processing more quickly. Models and datasets are growing at incredible rates and large systems that are not seeing performance degradation yet are all but guaranteed to see it in the near future.

Relatedly, a parallel computing environment like those running Apache Spark might see significant performance gains with a SiliconDisk. Beyond typical operations, a disk and network I/O heavy operation like shuffle could benefit as well. Spark shuffle places high demands on the I/O capabilities of its environment: disk I/O, network I/O, with dense serial and deserialization operations all occurring simultaneously.

Virtual environments by design involve heavy caching and that activity only increases when virtual machines (VMs) migrate. ATTO XstreamCORE® already speeds up data and VM migration in virtual environments, and with SiliconDisk added as a caching device, the process could occur faster than anyone can presently imagine.

While all of the above are specialized workloads, even the more traditional workloads are becoming complex behind the curtain. A typical marketing department will reference numerous data stores in its analysis of captured metrics; a film or media studio will soon have to begin the transition to 8K streaming video which will place phenomenal demands on architectures that can narrowly keep up with 4K workflows; data transactions of all kinds occurring across a university already need updating when file sizes are growing. SiliconDisk could solve the time-wasting, sometimes anxiety-provoking, stuttering and delays caused by storage and storage network inefficiencies.