ARM SERVER STANDARDIZATION

Similar documents
ServerReady and Open Standards Accelerating Delivery

Introduction to Standards based approach to Server

Standardized Firmware for ARMv8 based Volume Servers

Updates on Server Base System Architecture and Boot Requirements. Dong Wei

Hypervisors at Hyperscale

The Role UEFI Technologies Play in ARM Platform Architecture

Veloce2 the Enterprise Verification Platform. Simon Chen Emulation Business Development Director Mentor Graphics

Building the Ecosystem for ARM Servers

ODP Relationship to NFV. Bill Fischofer, LNG 31 October 2013

Zhang Tianfei. Rosen Xu

Arm Server Ready. Dong Wei

Maximizing heterogeneous system performance with ARM interconnect and CCIX

SmartNICs: Giving Rise To Smarter Offload at The Edge and In The Data Center

CCIX: a new coherent multichip interconnect for accelerated use cases

The Convey HC-2 Computer. Architectural Overview. Convey White Paper

Next Generation Enterprise Solutions from ARM

Post-K Development and Introducing DLU. Copyright 2017 FUJITSU LIMITED

Digitizer operating system support

UEFI in Arm Platform Architecture

Accelerating Data Centers Using NVMe and CUDA

OpenDataPlane (ODP) A Quick Introduction and Overview. Linaro Networking Group (LNG) Presented by Bill Fischofer.

System-on-Chip Architecture for Mobile Applications. Sabyasachi Dey

Bringing Intelligence to Enterprise Storage Drives

RapidIO.org Update.

FreeBSD on Cavium ThunderX System on a Chip

Post-K Supercomputer Overview. Copyright 2016 FUJITSU LIMITED

Cavium Introduces ThunderX : A 2.5 GHz, 48 Core Family of Workload Optimized Processors for Next Generation Data Center and Cloud Applications

Cisco HyperFlex HX220c M4 Node

SUSE Linux Entreprise Server for ARM

UEFI updates, Secure firmware and Secure Services on Arm

DynamIQ Processor Designs Using Cortex-A75 & Cortex-A55 for 5G Networks

ISR G2 and Service Ready Engine Smart Branch Vision

Netronome NFP: Theory of Operation

RapidIO.org Update. Mar RapidIO.org 1

Hardened Security in the Cloud Bob Doud, Sr. Director Marketing March, 2018

Building High Performance, Power Efficient Cortex and Mali systems with ARM CoreLink. Robert Kaye

ARM Server s Firmware Security

SCHOOL OF PHYSICAL, CHEMICAL AND APPLIED SCIENCES

PreBoot Provisioning Solutions with UEFI

Jumpstart your Production OpenStack Deployment with

Copyright 2017 Intel Corporation

Hardware Software Bring-Up Solutions for ARM v7/v8-based Designs. August 2015

Cisco HyperFlex HX220c M4 and HX220c M4 All Flash Nodes

Virtualization Benefits IBM Corporation

PCI Express x8 Single Port SFP+ 10 Gigabit Server Adapter (Intel 82599ES Based) Single-Port 10 Gigabit SFP+ Ethernet Server Adapters Provide Ultimate

Network Adapters. FS Network adapter are designed for data center, and provides flexible and scalable I/O solutions. 10G/25G/40G Ethernet Adapters

Introduction to Configuration. Chapter 4

Tutorial. Preparing for Stampede: Programming Heterogeneous Many-Core Supercomputers

Enabling and Optimizing MariaDB on Qualcomm Centriq 2400 Arm-based Servers

Storage Controller Considerations

Optimizing ARM SoC s with Carbon Performance Analysis Kits. ARM Technical Symposia, Fall 2014 Andy Ladd

FPGAs as Tools and Architectures at ETH Systems

RESTRUCTURING DPDK DEVICE-DRIVER FRAMEWORK

Intel Acceleration Stack for Intel Xeon CPU with FPGAs Version 1.2 Release Notes

DPDK Roadmap. Tim O Driscoll & Chris Wright Open Networking Summit 2017

CLICK TO EDIT MASTER TITLE STYLE. Click to edit Master text styles. Second level Third level Fourth level Fifth level

SAP High-Performance Analytic Appliance on the Cisco Unified Computing System

Offloaded Data Transfers (ODX) Virtual Fibre Channel for Hyper-V. Application storage support through SMB 3.0. Storage Spaces

Unify Virtual and Physical Networking with Cisco Virtual Interface Card

Red Hat Virtualization 4.1 Hardware Considerations for Implementing SR-IOV

IOmark- VM. HP MSA P2000 Test Report: VM a Test Report Date: 4, March

Verification Futures Nick Heaton, Distinguished Engineer, Cadence Design Systems

Distributed Data Infrastructures, Fall 2017, Chapter 2. Jussi Kangasharju

Red Hat Enterprise Virtualization 3.6

NFV Infrastructure for Media Data Center Applications

Towards Converged SmartNIC Architecture for Bare Metal & Public Clouds. Layong (Larry) Luo, Tencent TEG August 8, 2018

Enyx soft-hardware design services and development framework for FPGA & SoC

Early Software Development Through Emulation for a Complex SoC

How Might Recently Formed System Interconnect Consortia Affect PM? Doug Voigt, SNIA TC

Xen on ARM. Stefano Stabellini

Storage Controller Considerations

To EL2, and Beyond! connect.linaro.org. Optimizing the Design and Implementation of KVM/ARM

Cisco HyperFlex HX220c M4 and HX220c M4 All Flash Nodes

CISCO EXAM QUESTIONS & ANSWERS

HETEROGENEOUS SYSTEM ARCHITECTURE: PLATFORM FOR THE FUTURE

SoC Systeme ultra-schnell entwickeln mit Vivado und Visual System Integrator

Bringing Intelligence to Enterprise Storage Drives

BKK16-309B Enterprise Firmware - The gold standard and how to get there. Jeff Underhill

Design, Verification and Emulation of an Island-Based Network Flow Processor

Software Development Using Full System Simulation with Freescale QorIQ Communications Processors

Datacenter Java Developers Start your ARMv8 Engines! CON11179

Integrating CPU and GPU, The ARM Methodology. Edvard Sørgård, Senior Principal Graphics Architect, ARM Ian Rickards, Senior Product Manager, ARM

DRAM and Storage-Class Memory (SCM) Overview

OpenCAPI Technology. Myron Slota Speaker name, Title OpenCAPI Consortium Company/Organization Name. Join the Conversation #OpenPOWERSummit

REQUEST FOR PROPOSAL FOR PROCUREMENT OF

HPC over Cloud. July 16 th, SCENT HPC Summer GIST. SCENT (Super Computing CENTer) GIST (Gwangju Institute of Science & Technology)

Pactron FPGA Accelerated Computing Solutions

Copyright 2016 Xilinx

Designing, developing, debugging ARM Cortex-A and Cortex-M heterogeneous multi-processor systems

Is There Any Alternative To Your Enterprise UNIX Platform? Andrej Gursky PosAm TechDays EAST, March 2015

CMP Conference 20 th January Director of Business Development EMEA

Did I Just Do That on a Bunch of FPGAs?

SoftFlash: Programmable Storage in Future Data Centers Jae Do Researcher, Microsoft Research

Best Practices for Setting BIOS Parameters for Performance

Calxeda : RACK TRUMPS THE CHIP

Storage Controller Information

CS/ECE 217. GPU Architecture and Parallel Programming. Lecture 16: GPU within a computing system

Survey of ETSI NFV standardization documents BY ABHISHEK GUPTA FRIDAY GROUP MEETING FEBRUARY 26, 2016

Voltaire. Fast I/O for XEN using RDMA Technologies. The Grid Interconnect Company. April 2005 Yaron Haviv, Voltaire, CTO

Simulation Based Analysis and Debug of Heterogeneous Platforms

Transcription:

ARM SERVER STANDARDIZATION (and a general update on some happenings at Red Hat) Jon Masters, Chief ARM Architect, Red Hat

6+ YEARS OF ARM AT RED HAT Red Hat ARM Team formed in March 2011 Bootstrapped ARMv8 architecture for RHEL (and Fedora, and CentOS, etc.) Co-created ARM server industry standards along with Linaro Enterprise Group (LEG) Collaborated with every ARM server chip semiconductor during design and verification Shipped Red Hat Enterprise Linux Server for ARM (Development Preview) Watch this space

Source: By Dr-text - Datacenter, CC BY-SA 3.0 https://commons.wikimedia.org/wiki/file:datacenter-mivitec-munich.jpg

Source: By Alan Chia - Lego Color Bricks, CC BY-SA 2.0 https://commons.wikimedia.org/w/index.php?curid=6068229

Source: By Arp - https://commons.wikimedia.org/wiki/file:maxscan_oe509_collage.jpg

Source: http://www.thecompaqstory.com ( Open, by Rod Canion)

SERVER BASE SYSTEM ARCHITECTURE

SYSTEM ON CHIP CPU Core Cluster Interconnect

SYSTEM ON CHIP Memory Controller

SYSTEM ON CHIP Serdes (PCIe, Fabric, SATA )

SYSTEM ON CHIP Workload Customized IP Compression Engine

SILICON COMPLIANCE Adherence to the ARM Server Base System Architecture (SBSA) Core compliance is (well) handled by ARM s architecture validation Uncore components sometimes need work (PCIe IP example) PCIe quirks are common across arches but need upstreaming Always design with backward compatibility in mind

SERVER BASE BOOT REQUIREMENTS

FIRMWARE COMPLIANCE Real servers are built using standard platforms. In the ARM community, we defined the ARM Server Base Boot Requirements (SBBR). We use UEFI and ACPI because they offer the necessary pieces for platform abstraction required in general purpose OSes. There is one driver for the Real Time Clock (RTC) in RHEL We have defined interfaces for provisioning ARMv8 servers (not your ARMv8 server), and for upgrading firmware easily over time

VALIDATION AND COMPLIANCE

OUT OF BOX EXPERIENCE Having a high performance server core isn t the first step to world domination Things need to just work in the way that users already expect them to Unbox, plug into rack, provision via the network using standard tooling Practice good hygiene by giving users a boring experience Use the Linaro Enterprise Reference Platform, other examples as a base

WHAT NOT TO DO Don t ship some hack with your custom hacked up distro and kernel of the day Don t ship a special OS for ARMv8 server X that has no upgrade path By all means pre-install an example, but make sure it isn t a hack (if it says it is RHEL it needs to actually be RHEL) If you are unable to avoid a hack, have a good plan and work it through upstream

Source: By Rob Young, Bluff Sign Post, CC BY-SA 2.0 https://commons.wikimedia.org/wiki/file:bluff_sign_post,_new_zealand_(870336830).jpg

WORKLOAD OFFLOAD ACCELERATORS Traditional compute tasks replaced with CPU managed data movement to dedicated or reconfigurable accelerators ARMv8 servers integrate many workload accelerators including network packet processors, compression engines, and security/crypto processors Use otherwise dark silicon more effectively (e.g. phone all day battery) Source: Cavium Networks (ThunderX)

RECONFIGURABLE COMPUTING FPGAs allow for custom logic to be tailored to a specific VM, attached onchip/package, interconnect (e.g. CCIX) Logically coherent or non-coherent with regard to host system memory Coherent is easier to program (e.g. CUDA) but requires cache snoop hw Non-coherent device memory access using OS DMA primitives Unified Memory (CUDA) (transparent migration), memory hierarchy (OpenCL)

RECONFIGURABLE COMPUTING Red Hat Trilby is a simple Red Hat PCIe FPGA platform developed for research Implements a trivial search algorithm (not for high performance) Uses hand-written Verilog, a custom DMA engine, and a custom drivers Interfacing to Apache Spark is possible through a native (JNI) library etc. Red Hat Trilby PCIe-attached FPGA

ARM SERVER STANDARDIZATION (and a general update on some happenings at Red Hat) Jon Masters, Chief ARM Architect, Red Hat

6+ YEARS OF ARM AT RED HAT Red Hat ARM Team formed in March 2011 Bootstrapped ARMv8 architecture for RHEL (and Fedora, and CentOS, etc.) Co-created ARM server industry standards along with Linaro Enterprise Group (LEG) Collaborated with every ARM server chip semiconductor during design and verification Shipped Red Hat Enterprise Linux Server for ARM (Development Preview) Watch this space

Source: By Dr-text - Datacenter, CC BY-SA 3.0 https://commons.wikimedia.org/wiki/file:datacenter-mivitec-munich.jpg

Source: By Alan Chia - Lego Color Bricks, CC BY-SA 2.0 https://commons.wikimedia.org/w/index.php?curid=6068229

Source: By Arp - https://commons.wikimedia.org/wiki/file:maxscan_oe509_collage.jpg

Source: http://www.thecompaqstory.com ( Open, by Rod Canion)

SERVER BASE SYSTEM ARCHITECTURE

SYSTEM ON CHIP CPU Core Cluster Interconnect

SYSTEM ON CHIP Memory Controller

SYSTEM ON CHIP Serdes (PCIe, Fabric, SATA )

SYSTEM ON CHIP Workload Customized IP Compression Engine

SILICON COMPLIANCE Adherence to the ARM Server Base System Architecture (SBSA) Core compliance is (well) handled by ARM s architecture validation Uncore components sometimes need work (PCIe IP example) PCIe quirks are common across arches but need upstreaming Always design with backward compatibility in mind

SERVER BASE BOOT REQUIREMENTS

FIRMWARE COMPLIANCE Real servers are built using standard platforms. In the ARM community, we defined the ARM Server Base Boot Requirements (SBBR). We use UEFI and ACPI because they offer the necessary pieces for platform abstraction required in general purpose OSes. There is one driver for the Real Time Clock (RTC) in RHEL We have defined interfaces for provisioning ARMv8 servers (not your ARMv8 server), and for upgrading firmware easily over time

VALIDATION AND COMPLIANCE

OUT OF BOX EXPERIENCE Having a high performance server core isn t the first step to world domination Things need to just work in the way that users already expect them to Unbox, plug into rack, provision via the network using standard tooling Practice good hygiene by giving users a boring experience Use the Linaro Enterprise Reference Platform, other examples as a base

WHAT NOT TO DO Don t ship some hack with your custom hacked up distro and kernel of the day Don t ship a special OS for ARMv8 server X that has no upgrade path By all means pre-install an example, but make sure it isn t a hack (if it says it is RHEL it needs to actually be RHEL) If you are unable to avoid a hack, have a good plan and work it through upstream

Source: By Rob Young, Bluff Sign Post, CC BY-SA 2.0 https://commons.wikimedia.org/wiki/file:bluff_sign_post,_new_zealand_(870336830).jpg

WORKLOAD OFFLOAD ACCELERATORS Traditional compute tasks replaced with CPU managed data movement to dedicated or reconfigurable accelerators ARMv8 servers integrate many workload accelerators including network packet processors, compression engines, and security/crypto processors Use otherwise dark silicon more effectively (e.g. phone all day battery) Source: Cavium Networks (ThunderX)

RECONFIGURABLE COMPUTING FPGAs allow for custom logic to be tailored to a specific VM, attached onchip/package, interconnect (e.g. CCIX) Logically coherent or non-coherent with regard to host system memory Coherent is easier to program (e.g. CUDA) but requires cache snoop hw Non-coherent device memory access using OS DMA primitives Unified Memory (CUDA) (transparent migration), memory hierarchy (OpenCL)

RECONFIGURABLE COMPUTING Red Hat Trilby is a simple Red Hat PCIe FPGA platform developed for research Implements a trivial search algorithm (not for high performance) Uses hand-written Verilog, a custom DMA engine, and a custom drivers Interfacing to Apache Spark is possible through a native (JNI) library etc. Red Hat Trilby PCIe-attached FPGA