Big Data—unstructured data coming to organizations in unprecedented volume, variety, and velocity—poses tremendous challenges to data center speed and efficiency. The use of Big Data is widespread in business and government, driving economic activity, innovation and growth. New applications for business analytics continue to emerge. Big Data problems require light-weight processing of huge volumes of data.
Xockets™ offers a breakthrough acceleration architecture for these problems, preserving compatibility with established commodity x86 architectures and software. Xockets builds appliances using our hardware and software co-processors, integrated into commodity server platforms, to enable dramatic acceleration of big data analytics applications.
Unstructured data tools such as Map-Reduce, Graphs, and ETL distribute analytic calculations in a standard, structured data center. Xockets appliance works with and distributes among these rack servers to produce a dense, accelerated data center at much lower power.
Different applications use storage differently. Xockets software-defined storage infrastructure flexibly adapts its highest-density, highest IO storage platform: 72 TB of Application optimized Flash at 100’s of Gbps in a 1U geometry.
Xockets solutions can connect in excess of 100 thousand wimpy cores, each tightly coupled to software-defined infrastructure, petabytes of flash, and an abundance of programmable logic. Normally occupying a full data center, this Xockets deployment would require only four racks.
Unstructured programming and analytics frameworks, like Hadoop, expose the architectural deficiencies of conventional x86-based servers. In the face of huge data working sets and light-weight computation, Xockets offers a cooperative-compute acceleration architecture, preserving compatibility with established commodity x86 architectures and software.
Xockets is changing the economics about which data can be kept available for queries. Problem analysis which would previously have required a full rack can now be performed on a 1 or 2RU server, using much less power.
How is this possible? Xockets has built co-processors which, when integrated in a conventional server running conventional x86 CPUs, accelerate light-weight processes and remove contention for shared processing, memory, and memory bandwidth. Data is managed as it moves it through a network of highly connected processors that can handle the data at a power level appropriate to the computation. Instructions and their data are joined in an architecture in which an instruction stream attracts its data. The Xockets computational memory fabric enables solutions at a fraction of the server and power resources required for the standard approaches.
Parin has been obsessed with computer architecture since he was 13 years old. He has over fifteen years of experience at start-ups Exponential, Chromatic Research, Photuris, and Catenary as well as at public companies ATI and Fujitsu. He has worked in circuit design and architecture, developing memory, CPUs, GPUs, and NPUs. Most recently, Parin was a researcher at Stanford and holds a doctorate in theoretical physics.
Stephen has over twenty years of experience in operating system and system architecture, with focus on scalable infrastructure, distributed systems and concurrent methods. A Distinguished Engineer at Cisco Systems for years, Stephen architected the original distributed infrastructure at the heart of Cisco's core CRS router and has since led a project for Cisco's run-to-completion IOS to discover concurrency conflicts at runtime.
Daniel brings 18 years of experience in software development and over 10 years in management. Prior to joining Xockets, he was Sr. Manager of Engineering at Cisco Systems. Twice recipient of Cisco's Pioneer Award, he is passionate about promoting and productizing innovation. Daniel has a BS in EE from MIT and a MS in EE from Stanford University.
Xockets builds hardware and software acceleration into appliances for Big Data processing. Traffic management, high speed memory and dense computational capacity enables these solutions at a fraction of the server and power resources required for the standard approaches.
Xockets was formed in 2012, and graduated from the Alchemist accelerator in January 2013. We are based in the San Francisco Bay Area, conveniently located a 15 minute walk away from the San Jose CalTrain station, with great public transport links to the Capitol Corridor and ACE commuter trains.
One Almaden Blvd,
201 Mission Street,