arrow
Return

BlastFunction: A Full-stack Framework Bringing FPGA Hardware Acceleration to Cloud-native Applications

delete2022-01-11
delete6
PRE
AI
A
Andrea Damiani *
G
Giorgia Fiscaletti
M
Marco Bacis
R
Rolando Brondolin
M
Marco D. Santambrogio
DOI:10.1145/3472958delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
Cloud-native is the umbrella adjective describing the standard approach for developing applications that exploit cloud infrastructures' scalability and elasticity at their best. As the application complexity and user-bases grow, designing for performance becomes a first-class engineering concern. As an answer to these needs, heterogeneous computing platforms gained widespread attention as powerful tools to continue meeting SLAB for compute-intensive cloud-native workloads. We propose BlastFunction, an FPGA-as-a-Service full-stack framework to ease FPGAs' adoption for cloud-native workloads, integrating with the vast spectrum of fundamental cloud models. At the IaaS level, BlastFunction time-shares FPGA-based accelerators to provide multitenant access to accelerated resources without any code rewriting. At the PaaS level, BlastFunction accelerates functionalities leveraging the serverless model and scales functions proactively, depending on the workload's performance. Further lowering the FPGAs' adoption barrier, an accelerators' registry hosts accelerated functions ready to be used within cloud-native applications, bringing the simplicity of a SaaS-like approach to the developers. After an extensive experimental campaign against state-of-the-art cloud scenarios, we show how BlastFunction leads to higher performance metrics (utilization and throughput) against native execution, with minimal latency and overhead differences. Moreover, the scaling scheme we propose outperforms the main serverless autoscaling algorithms in workload performance and scaling operation amount.
Keywords:
Field Programmable Gate Arrays (FPGAs)
cloud-native
time-sharing
autoscaling
hardware acceleration

Journal

ACM Transactions on Reconfigurable Technology and Systems cover
ACM Transactions on Reconfigurable Technology and Systems
IF:
2.8
Papers:
598
Citations:
810

Organization

P
Polytechnic University of Milan
Scholars:
2.0W
Papers: 1.8W
Citations: 24
Cited Papers

Cited Papers

Linkage map positions and allelic diversity of two Mal d 3 (non-specific lipid transfer protein) genes in the cultivated apple (Malus domestica)
err2005-01-13
err0
PREAI
errZ. S. Gao; W. E. van de Weg; J. G. Schaart; I. M. van der Meer; L. Kodde; M. Laimer; H. Breiteneder; K. Hoffmann-Sommergruber; L. J. W. J. Gilissen
errShare
errSave
FPGA Resource Pooling in Cloud Computing
err2021-04-01
err25
errOAAI
errZhu, Zhuangdi; Liu, Alex X.; Zhang, Fan; Chen, Fei
errShare
errSave
Anisotropic Water-Mediated Proton Conductivity in Large Iron(II) Metal–Organic Framework Single Crystals for Proton-Exchange Membrane Fuel Cells
err2018-12-18
err0
PREAI
errHana Bunzen; Ali Javed; Danielle Klawinski; Anton Lamp; Maciej Grzywa; Andreas Kalytta-Mewes; Michael Tiemann; Hans-Albrecht Krug von Nidda; Thorsten Wagner; Dirk Volkmer
errShare
errSave
The effect of repeated testing of pharmacy calculations and drug knowledge to improve knowledge retention in pharmacy students
err2018-12-01
err0
PREAI
errAdeola O. Coker; Kathleen A. Lusk; David F. Maize; Sushma Ramsinghani; Regina A. Tabor; Elizabeth A. Yablonski; Alejandra Zertuche
errShare
errSave
Shadowlands and dark corners
err2018-06-25
err0
errOAAI
errAnn H. Kelly; Almudena Marí Sáez
errShare
errSave
researcher View more