Browse Prior Art Database

Hardware accelerator mobility in virtualized environments

IP.com Disclosure Number: IPCOM000247760D
Publication Date: 2016-Oct-06
Document File: 5 page(s) / 65K

Publishing Venue

The IP.com Prior Art Database

Abstract

Hardware accelerator mobility in virtualized environments

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 40% of the total text.

Page 01 of 5

Hardware accelerator mobility in virtualized environments

Acronyms Used:


LPM - Live Partition Mobility
CPU - Central Processing Unit
SRIOV - Single Root IO Virtualization
CAPI - Coherent Accelerator Processor Interface FPGA - Field Programmable Gate Array
AFUs -
Accelerated Function Units

Novelty:

This article provides a framework for migrating hardware accelerator functions used by a workload to another physical server during partition migration.

Current Problem:

In computing, hardware acceleration is the use of computer hardware to perform some functions faster than what is possible in software running on the general-purpose CPU. These accelerators are dedicated fixed-function blocks designed to perform a single computationally intensive task over and over. They offload the main processor, allowing it to do general-purpose tasks that have little regularity in structure.

Although hardware acceleration is used widely in enterprise computing, it is not widely used in virtualized environments. Technologies like SRIOV are bringing in ways to mitigate some of the issues that are blocking the deployment of RC devices.

The problem addressed by this article is the mobility of hardware accelerators in AIX LPM/ VMware vSphere like partition mobility scenarios.

Solution:

This article provides a framework for migrating hardware accelerator functions used by a workload to another physical server during partition migration. Although the examples cited in this article describes AIX specific technologies like LPM or CAPI, this is applicable to any partition migration scenarios in data-centers or cluster computing environments.

Live Partition Mobility allows you to migrate partitions that are running AIX and Linux operating systems and their hosted applications from one physical server to another without disrupting the infrastructure services. The migration operation,which takes just a few seconds, maintains complete system transactional integrity. The migration transfers the entire system environment, including processor state, memory, attached virtual devices, and connected users.

1


Page 02 of 5

Figure 1

Current partition mobility functionality does not support migration of accelerator functions (hardware accelerators) used by the workload. Hardware accelerator are generally available as on-chip processor accelerators, co-processors (functionality is not present in the CPU die, but as a direct attachment to CPU e.g. Intel Xeon FPGA device) or as a IO attachment (eg thorough PCI cards). The current scope of this article is applicable to FPGA accelerators and hence cover the latter two scenarios (CPU attach or IO attach).

In the above case (refer Figure 1), the workload in System #1 might be using an Accelerated Function Unit (AFU) present in its local CAPI device (PCI card). With the proposed feature, the migrated partition in System #2 will be able to use the same AFU if CAPI card is present in System#2 (with logic device space for AFU programming...