Accelerating The Cloud with Heterogeneous Computing

将多个CPU和GPU联合收割机组合在一个芯片上的异构多处理器将在市场上变得司空见惯。正如最近从高性能计算社区看到的那样，利用GPU可以产生几个数量级的性能提升。通过与AMD合作伙伴的合作，我们将研究如何使用GPU加速来大大加快虚拟化中心的云管理任务。这只有在GPU在片上移动的情况下才成为可能，因为PCIe总线上的延迟太大，无法在任何给定点对系统状态做出快速，明智的决定。该提案研究了可以从GPU加速中受益的云管理任务。它还研究了与如何在多租户系统中管理此硬件相关的坚韧问题。GPU加速系统任务将降低云提供商的成本，并使他们能够提供增值服务，例如自动病毒检查，从而在竞争激烈的市场中实现服务差异化。如果成功，拟议的研究将为AMD在数据中心市场提供竞争优势，这是其业务的核心领域，但传统上与其GPU技术的优势无关。

Heterogeneous multiprocessors that combine multiple CPUs and GPUs on a single die are poised to become commonplace in the market. As seen recently from the high performance computing community, leveraging a GPU can yield performance increases of several orders of magnitude. In collaboration with our AMD partner, we will research the use of GPU acceleration to greatly speed up cloud management tasks in virtualized datacenters. This is only becoming possible now that the GPU is moving on-chip, since the latency across the PCIe bus was too great to make fast, informed decisions about the state of a system at any given point. This proposal researches cloud management tasks that can greatly benefit from GPU acceleration. It also investigates the tough questions related to how to manage this hardware in a multi-tenant system. GPU acceleration of systems tasks would reduce costs for cloud providers and make it possible for them to offer value-add services, such as automatic virus checking, that allow for service differentiation in a fiercely competitive market. If successful, the proposed research will provide AMD a competitive advantage in the datacenter market, a domain that is a core to its business, but that has not been traditionally associated with its strength in GPU technology.