Intelligent Infrastructure Lab

Goal

Build and manage Intelligent Infrastructure fabrics (compute, storage, photonics, sensors, etc.) for next-generation IT, Cloud, and embedded solutions with dramatically improved scale, performance, reliability and cost of ownership. In particular:

  • Next Generation Data Centers: Build compute fabrics for next-generation IT solutions utilizing a cross-layer, inter-disciplinary approach across conventional divisions – of CPU, system, and data center; compute, storage, networking, and packaging; hardware, firmware, systems, and software; over performance, power, availability, and manageability; with cost-effectiveness and agility.
  • CeNSE: Create the mathematical and physical foundations for the technologies that will form a new information ecosystem, the Central Nervous System for the Earth (CeNSE), consisting of a trillion nanoscale sensors and actuators embedded in the environment and connected via an array of networks with computing systems, software and services to exchange their information among analysis engines, storage systems and end users.
  • Intelligent Storage: Research and build the platforms and services that will provide the next generation of storage and information services for enterprise and user-controlled information. The platforms will be scalable, distributed, and readily extended, and will provide unequalled reliability, performance, scalability cost effectiveness and manageability. By combining processing power and storage, we aim to enable the next generation of information management capabilities.
  • Photonics: Increase the performance per unit cost of commercial interconnects by a factor of 2 (in 2012) to 20 (in 2017) above that available from currently-planned electronic interconnects over distance scales of 1 mm–100 m.

Impact

Dramatically improved performance (>10x), scale (>100x), cost of ownership (<10x) and reliability (6+ 9’s) of emerging cloud/data-centric and traditional data centers, and sensors-based solutions, enabling new kind of applications and services.

  • Next Generation Data Centers: We expect our research to enable “cloud” end-to-end data center solutions with significant (5-10X) improvements in performance-per-cost compared to the state-of-the-art and a top 500 High Performance Computing demonstrator whose performance significantly (15X-20X) leapfrogs that of previous systems.
  • CeNSE: Revolutionize human interaction with the earth as profoundly as the Internet has revolutionized personal and business interactions.
  • Intelligent Storage: We expect our research to enable new cloud storage services, which significantly extend the state of the art in reliability and cost-effectiveness. In addition, we will continue to deliver significant cost/performance advantages to our products in areas such as deduplication. Finally, we will enable HP to become a key player in the information management and governance space, particularly for unstructured data.
  • Photonics: Establish a highly scalable and cost-effective “Moore's Law” for information bandwidth at all levels of granularity from the chip to the enterprise in HP systems.

Research

Overall, the Lab will focus on research to:

  • Dramatically improve the manageability, power, and availability of future computing systems.
  • Significantly improve performance/cost of systems by leveraging future technology (e.g., nanophotonics, memristors, etc.).
  • Build new scalable, reliable and cost-effective cloud storage systems and develop systems to deal with the information explosion of unstructured data.
  • Advance coding, compression, and other information-theoretic technologies in support of next-generation memories, massive sensor networks, storage systems, and high-reliability computing.

In the area of the Next Generation Data Centers, the Lab will focus on research to:

  • Develop effective system architectures and virtualization/management solutions for the exascale era.
  • Develop new co-designed hardware/software solutions to leverage emerging non-volatile memories, optics, and system integration in multi-cores.
  • Develop effective parallel programming techniques and software tools for large clusters of many-core systems.

In the area of CeNSE, the Lab will focus on research to:

  • Dramatically improve the state of the art in the cost, size, power efficiency, accuracy, and sensitivity of embedded sensors and actuators.
  • Study applications of nonlinear optics at the nanoscale (such as surface Raman scattering using plasmonic structures) to provide novel chemical sensors and analytics.

In the area of information theory, the Lab will focus on research on:

  • Coding technologies for ensuring the reliability of computer memory systems based on emerging memory devices, with an emphasis on memristor-based technologies
  • Energy-aware data compression with application to CeNSE, aiming at algorithms optimizing the tradeoff between power consumption for computation, storage, and/or communication, sensor data fidelity, and compressed data rate.
  • Coding for multi-user communication, with the goal of designing efficient and reliable communication fabrics (including codes and protocols) for large sensor networks.

In the area of intelligent storage, the Lab will focus on research on:

  • Scalable storage for “the cloud” (enterprise quality of service at consumer price point).
  • Content-aware storage (content and context analysis to enrich stored items and metadata).
  • Automating information management (leveraging content and context to automate storage and information management).

In the area of photonics, the Lab will focus on research to:

  • Develop photonic interconnects enabled by extremely efficient and inexpensive optical sources, modulators and detectors; develop advanced optomechanical packaging technologies for low cost, high bandwidth, high density optical interconnections.
  • Create new classes of optical technologies that rely on advanced physics of new materials; use CMOS-compatible nanoscale linear and nonlinear optics to dramatically increase interconnect bandwidth density at costs far below those of copper interconnects.
  • Enable new systems, computer processors, and switch architectures through large-scale integrated photonics; and explore new approaches to classical computing based on quantum technologies engineered for ultra-low-energy operation.

Director: Norman Jouppi