NanoStreams co-designs a micro-server architecture and a software stack that address the unique challenges of hybrid transactional-analytical workloads, which are encountered by emerging applications of real-time big-data analytics. To this end, NanoStreams brings together embedded system design principles, application-specific compiler technology, and HPC software practices.

The processor technology that underpins the NanoStreams micro-server is an amalgam of RISC cores and nano-cores, a new class of programmable, custom accelerators. Novel automatic compiler generation and parameterisation technology enables low-effort programming and integration of nano-cores into application-specific, many-core accelerators. The proposed heterogeneous Analytics-on-Chip processor forms the backbone of the NanoStreams micro-server, which further leverages a hybrid DRAM-PCRAM memory system and a non-cache-coherent scale-out architecture to achieve extreme energy-efficiency.

The software stack of the NanoStreams micro-server is rooted in domain-specific languages for analytical queries, which the project implements with a streaming dataflow execution model. The language runtime system uses real-time scheduling, performance isolation techniques and region-based locality management to minimise latency on the transactional path and maximise throughput on the analytical path. NanoStreams virtualises lightweight PCRAM-based persistent memory, for direct user access and locality optimisation.

The project will deliver a real-silicon prototype, based on the Xilinx Zynq platform and ARM-Linux. The quantitative objective of NanoStreams, in comparison with contemporary HPC servers, is to reduce analytical response time of commercial in-memory databases by at least 30%, while sustaining transactional throughput and improving system energy-efficiency and programmability. NanoStreams will demonstrate these advances with industry-standard workloads and four real-world case studies.

NanoStreams brings together a consortium of two academic institutions (Queen’s University of Belfast, FORTH-ICS), three technology-providing SMEs (Analytics Engines Ltd., ACE, Neueda), one global enterprise developing holistic solutions (hardware and software) for high-performance real-time data analytics (IBM) and one investment bank (Credit Suisse). The NanoStreams consortium has a balanced expertise in HPC and embedded systems. The five industrial partners present excellent opportunities for commercialisation of the project results.

Follow project progress on the NanoStreams website here.