Intelligent infrastructure

Research opportunities and challenges

Modern enterprises create and process information at ever-expanding volume and accelerating rates. Today’s infrastructure, built on old protocols, is strained by enterprise requirements to support bandwidth-hungry applications, mobility, security, and manageability. Infrastructure elements requiring innovation include:

• Computing architectures based on parallel computing, in which every processor chip has multiple CPU cores, requiring a new generation of software applications and firmware to exploit the new technology.
• Networks must scale to support datacenter growth. This drives demand for a robust, programmable wired and wireless network platform that makes the introduction of new features quick, easy and cost-effective.
• Storage must become capable of self-management and meet all the demands of the next generation enterprise computing infrastructure.

HP Labs research in Intelligent Infrastructure includes five large projects:

Next Generation Datacenters

The explosion of data that enterprises, consumers and sensors continuously generate, store, search and mine requires new thinking about the way in which the datacenters processing this information have to be designed, managed and programmed. Our end goal is to build a scalable, power- and cost-efficient, automated and programmable datacenter for the "data-centric" society. To achieve this, we have structured a research plan that ties together novel computing platforms, photonic interconnects, scalable management software, and programming platforms.

Networking

Networking’s grand vision is to deliver network innovations an order of magnitude faster than what is possible today and at a tenth of what it costs today. By developing a programmable network fabric over commodity devices and a programming environment for the fabric, we enable the rapid, economical introduction of new network features and offer greater control to enterprise and data center customers for their network. We are building and deploying a rapidly-customizable, programmable data center and enterprise network, and deploying new applications on these networks.

Next Generation Scalable Storage

The goal of our research is to design the storage platform for cloud computing. Our approach is to exploit a fundamental tradeoff between the consistency, availability, and network-partition-tolerance of distributed systems. Our system modifies the consistency models to offer much higher availability and the ability to update data even when there are network failures, at a low cost. Attributes of our system include scalability, ease of management, including capacity planning, automatic data reorganization, automatic tuning, and per-client QoS.

Nonvolatile Memory and Storage

The Memristive memory program’s goal is to create low power, high speed, ultra-high density, low cost and nonvolatile universal memory and storage solutions that eventually replace Flash, hard disks, DRAM, SRAM. Our research path is to continuously invent and improve new memristive materials, device structures, hybrid CMOS/memristor circuits, storage and logic architectures, and information theoretic strategies tailored to specific applications. We develop significant intellectual property and work with external manufacturing partners for development and commercialization of essential components.

CeNSE

CeNSE envisions trillions of interconnected inexpensive sensors measuring and processing all sorts of data in real time, to improve safety, sustainability, and security of people and businesses. Our research lays the conceptual groundwork for the communication fabric of a CeNSE-scale sensor network, with an emphasis on the wireless component. Areas of focus include the tools of network information theory, allowing for new architectures, protocols, and codes that go well beyond those deployed in state-of-the art networks today.