Our research slogan is: The right data, in the right place,
all the time.
This is not primarily about making better disk drives, or better disk arrays (although technical progress in both areas will continue to be important). This vision is about turning the data storage system as a whole into a truly ubiquitous, dependable service - a utility. This service will ensure that your data is made available to you, wherever you need it, whenever, and meeting your quality of service requirements.
Performance, security, capacity, availability, and data
protection guarantees will be met - both for current needs and for the
rate at which they change. It will do so both inside large-scale data
centers, and across the globe. Data will be migrated to wherever it
is needed to meet the customer's access requirements.
The storage utility will eliminate the error prone manual processes
of system design, configuration, monitoring and data migration by
automating them, developing a true lights-out storage management
system. In turn, this will help to alleviate the critical shortage of
trained system administrators, and drastically reduce response times
to changes in workload (demand).
Our business emphasis is on the corporate data processing needs,
and the storage service providers (SSPs and other xSPs) that provide
them. In the future, we believe that business will no longer purchase
storage boxes, but will instead purchase access to a storage utility.
They will pay for meeting their business needs, rather than a kit of
parts they must assemble themselves. And we believe that this future
is not far away. [Come help us make it happen!]
federated storage system - This project is focused on creating the global storage mesh from commodity components. The key question we're addressing is how to efficiently connect hundreds of thousands of storage "bricks" to provide a storage system that infinitely scales in performance, capacity and availability, while also being self-managing, self-healing and self-configuring.
self-managing storage - Researchers are working to create the self-managing storage system, allowing people to specify high-level goals and leaving the system to implement the details. As a result, the amount of storage managed by a single administrator will increase dramatically. We are inventing a set of technologies that can automatically manage an exabyte-size storage system spread across a wide area network.
scalable storage security - The global storage mesh extends across the public Internet, and the servers and storage devices will be hosted in service providers' data centers. As such, these are fundamentally untrusted environments, and so securing the data is a key research issue. There are two fundamental questions that have to be answered in securing these untrusted environments, namely, encrypting on the wire or on the storage and cryptographic key distribution and management.
architectures for new storage technologies - In this project, researchers are examining how the traditional architectures of high-end storage systems may incorporate new nonvolatile storage technologies: Atomic Resolution Storage (ARS) and Magnetoresistive RAM (MRAM).
global file service - The goal of this program is to design a file service that enables file placement and access on a global scale. To do that, we're working on two fundamental research problems;
- How to aggregate and virtualize resources on a global scale (thousands of servers, petabytes of storage) and how to facilitate on-line reconfiguration to meet fluctuating demand.
- How to perform automatic and transparent file placement and replication across the available resources, according to factors such as locality of access, resource utilization or content type.
To do our work, we've developed a suite of software tools. Several of which we are willing to make available to university researchers. We have made some of our tools such as Lintel and DataSeries available as open-source.
things we're looking to collaborate on
We're a relatively small group, and keep on coming up with what (we
think are) neat ideas that we don't have time to explore as far as
we'd like to. However, we'd be delighted to work with somebody else
to help make them become reality. In many cases, these would make
fine masters-level projects; in some cases, they're worthy of a PhD
(or two!). If you think you might be interested in working with us, please
contact our manager Alistair Veitch (alistair.veitch AT hp.com).