Project Details
Description
Handheld devices such as cell phones, tablets and wearables, are gaining
rapid adoption in a ubiquitous world. Correspondingly, there is an
explosion of very demanding and interactive applications such as media
streaming, interactive games, video conferencing and social networking,
that users are deploying on such devices. Despite tremendous
improvements in the energy efficiencies of individual components found
in these devices, piece-meal optimizations of hardware components are
inadequate to accommodate the demanding needs of these applications that
employ multiple CPU cores, accelerators, memory systems, system
peripherals and sensors, concurrently in very sophisticated ways. Lack
of coordination among system resources leads to low resource utilization
and poor system-level energy efficiencies when running such
applications. Rather than throw more hardware at this problem, this
project proposes VIADUCT, which allocates and orchestrates system-level
resources to boost parallelism and resource utilization for these
platforms in a more energy-efficient fashion.
Recognizing that many of these applications periodically process frames
of data that need to flow in a pipelined manner through several hardware
components within soft real-time bounds, VIADUCT creates a virtual
channel per flow of such frames in an application. The channel holds
resources allocated across all the hardware components, with runtime
coordination to ensure seamless frame movements without the conventional
interface inefficiencies across the components. Fine-grain scheduling
and resource management leverages knowledge of high level semantics to
provide Quality-of-Service (QoS) for each flow in the most
energy-efficient fashion. The specific research contributions are in
hardware interfaces, mechanisms and policies to chain the hardware
components for creating a virtual channel for a flow, architecting the
memory system as an efficient conduit to sustain this flow, and
designing the system to sustain several concurrent flows. The project
aims to provide a holistic evaluation testbed for such platforms amongst
the broader research community, and also aims to involve
under-represented groups in the research and enhance Penn State?s
curriculum in related topics.
Status | Finished |
---|---|
Effective start/end date | 8/1/15 → 7/31/20 |
Funding
- National Science Foundation: $499,998.00