NGD Systems to Showcase FAISS Running on In-Situ based NVMe SSDs at the 2018 OCP Summit, March 20-21, 2018
Computational storage company NGD Systems, Inc. announced today that it will showcase the potential of computational storage by augmenting the use of the Facebook Artificial Intelligence Similarity Search (FAISS) in a proof of concept (POC) around a distributed implementation of Approximate Nearest neighbor (ANN) search, where NGD’s computational storage Catalina-2 (CAT-2) In-Situ Processing NVMe SSDs are leveraged to do fast, scalable search of nearest neighbors for very large image datasets. The showcase will occur in the NGD Systems Booth (Booth B35) at the Open Compute Project (OCP) Summit on March 20-21, 2018 at the San Jose Convention Center, San Jose, Calif. The capabilities of In-Situ based improvements to FAISS show that real, meaningful workloads can run through in-device processors contained in NGD Systems computational storage devices and can accelerate performance at a significantly lower cost than can be realized with conventional server technologies.
“IT infrastructure is amid a massive transformation towards distributed architectures that optimize the usage and scaling of both compute and capacity. Users are eager for innovative solutions that help deliver immediate results like the In-Situ Processing SSDs from NGD Systems,” said Stu Miniman, Sr Analyst at Wikibon. “NGD Systems’ showcase of real world improvements to applications like FAISS continues to show that computational storage is no longer a nascent market space.”
FAISS is a library for efficient similarity search and clustering of dense vectors. It can quickly search for multimedia documents that are similar to each other, an ability that standard search approaches such as SQL databases cannot match. It contains algorithms that search in set of vectors of any size, up to ones that possibly do not fit in RAM. It is built as a C++ library with a full Python wrapper and was developed by the Facebook AI Research team.
“Organizations that deal with very large data sets have always had the challenge of deciding how to ingest significant amounts of data while getting results in real-time from that data,” said Scott Shadley, Vice President of Marketing at NGD Systems. “Our computational storage devices, embodied in our In-Situ NVMe SSDs, change the plan of attack for large data sets by providing in-device processing, simplifying the hardware designs over what was previously required. Our showcase of In-Situ Processing on a CAT-2 augmenting FAISS capabilities highlights the impact that computational storage has on complex workloads.”
NGD Systems, through its EMEA partner NVMeStorage, is also showcasing the potential of computational storage with a CAT-2 In-Situ Processing SSD running advanced application acceleration on the show floor. The product showcase will occur in the NGD Systems Booth (Booth R31) at CloudFest 2018 on March 12-15, 2018 at Europa-Park in Rust, Germany.
The NGD Systems In-Situ based computational storage SSDs utilize a patented breakthrough architecture which permits it to provide in-storage processing, high capacity, and low power consumption for a variety of target markets. The current CAT-2 family of drives also utilizes NGD Systems’ proprietary QoS technology to manage power, flash endurance, and latency characteristics, as well as an NGD Systems-proprietary advanced LDPC error correction capability to ensure complete protection of its contents.
Founded in 2013 with HQ in Irvine, Calif., NGD Systems is a venture-funded company focused on creation of new category of storage devices that brings computation to data. NGD has designed its advanced proprietary controller technology which deploys patented Elastic FTL algorithm and Advanced LDPC Engines to provide industry leading capacity scalability. The controller also deploys the patented In-Situ Processing technology to enable Computational Storage capability. The company is led by an executive team that helped drive and shape the flash storage industry, with decades in leadership positions with storage companies such as Western Digital, STEC, Memtech, and Micron.