Edge Computing With Layerscape

Joseph Byrne

May 2018

Company Public– NXP, the NXP logo, and NXP secure connections for a smarter world are trademarks of NXP B.V. All other product or service names are the property of their respective owners. © 2018 NXP B.V. Agenda

• Edge Computing Introduction • Layerscape Processors • Layerscape Software and EdgeScale • Edge Computing Applications • Conclusion

COMPANY PUBLIC 1 Introduction

COMPANY PUBLIC 2 Edge Computing Definition Data Goes • Inclusive to Processing − Computing near the source/sink of data − AKA moving computing to the data • Narrow − Applying cloud-computing techniques outside the data center ▪ Soft provisioning of compute, storage, networking ▪ Virtualization and containerization 10110101 ▪ Service-oriented architecture

▪ Orchestration Processing Goes to Data

COMPANY PUBLIC 3 Related Concepts

• Fog − For fog adherents, fog nodes are capable participating in distributed analytics − NXP does not distinguish between fog and edge

• Hybrid cloud − An IT term describing linking private and public clouds − Analogous to edge computing but unrelated

• Embedded processing − General term for computing done in a system without user-loaded software

• Internet of Things (IOT) − Narrow: embedded systems with internet connections − Expansive: any system or device accessible directly or indirectly via the internet − Complements edge computing

COMPANY PUBLIC 4 Edge Computing Is On Premises or in the Network

✓

✓ ✓

✓ ✓ ✓ ✓

IOT Gateway/Nexus MEC Premises Telco Network

MEC = multi-access (mobile) edge computing COMPANY PUBLIC 5 Edge Computing Evolutionary Stages

1. Precursor 2. Cloud Computing 3. Re-Localization 4. Local Cloud

Local Functions Added Functions in Cloud APIs Command Via Cloud Cloud Integrated Implemented and Control Computing Locally Locally

Pre-Edge Computing True Edge Computing

COMPANY PUBLIC 6 Edge Computing Topologies

• Dimension 2: Computational Topology

• Self-contained: Edge node does all computation for a specific machine or IoT endpoint

• Hub and spoke: One edge node services multiple machines/endpoint

• Peer-to-peer: Loads migrate among nodes with free capacity or the cloud

• Hierarchical: Edge node shares computation, example: − Cloud trains models − Endpoint classifies observations (e.g., recognizes objects) − Edge node decides on actions

COMPANY PUBLIC 7 Edge Computing Advantages Over Cloud Computing

Reduce Data Reduce Secure Data Transferred Latency Onsite

COMPANY PUBLIC 8 Edge (and Cloud) Computing Advantages Vs Traditional Embedded Abstract Scalable and Elastic Manageable

COMPANY PUBLIC 9 Major Edge Computing Players

OEMs/ODMs

Cloud Computing Telcos Companies

Store Store

Consumers & Businesses

COMPANY PUBLIC 10 Complications

Ecosystem Manageability Security

Abstract programming Cloud frameworks help Edge nodes must not be and deployment models manage applications, recruited to a thingbot demand a standard but what about far-flung army platform devices?

COMPANY PUBLIC 11 Agenda

• Edge Computing Introduction • Layerscape Processors • Layerscape Software and EdgeScale • Edge Computing Applications • Conclusion

COMPANY PUBLIC 12 Layerscape Processors

COMPANY PUBLIC 13 Generic Layerscape Block Diagram 1-16 CPUs

Trust Multiple Multiple Architecture Arm® V8 CPUs Arm® V8 CPUs SEC

Cache Integrated ECC DDR4/DDR3L DDR

GPU and LCD Other Controller Display Ethernet PCIe USB I/O DisplayPort PHY 4K Display

Ethernet Controllers Gen 2, 3, or 4 1, 2.5, 5, 10, 25, 100 Gbps COMPANY PUBLIC 14 A Broad and Scalable Edge Computing Portfolio

Video/image LS1028A LS1043A LS1046A LS1088A processing, • Cortex-A72 • Cortex-A53 • Cortex-A72 • Cortex-A53 large-scale • 2 cores • 2-4 cores • 2-4 cores • 4-8 cores analytics, TSN ethernet, • 1.6GHz • 1.6GHz • 1.8GHz • 1.6GHz gateway • 4-9W • 1/10G Ethernet, • 1/10 G Ethernet, • 1/10 G Ethernet, applications • Integrated TSN USB, PCI USB, PCI USB, PCI switch, GPU • 5-10W • 10-12W • 8-16W

Data acquisition, LS1012A LS1021A Large-scale LS2088A LX2160A analytics, • Cortex-A53 • Cortex-A7 video/image • Cortex-A72 • Cortex-A72 monitoring, • 1 core • 2 cores processing, • 4-8 cores • 8-16 cores remote control • 1GHz • 1.2GHz data • 2.0GHz • 2.2GHz aggregation, • 1-2W • 2W backhaul • 8 x 10GE • 10/25/40/100 GE • Ethernet, USB, PCI • Ethernet, USB, PCI • 20-35W • 31W

• All LS-series processors have rich set of IO – USB, PCIE, SATA, GPIO, I2C, SPI, UART • All support Trust Architecture for platform security • Support both embedded and PC Linux distros • Support industrial temperature ranges • Support long lifecycles COMPANY PUBLIC 15 One Package, Four SoC Options Arm Cortex-A72 Doubles Performance Vs Cortex-A53 4x A53 1.6 GHz LS1043 LS1046 4x A72 1.8 GHz 4.2 W Typical 8.5 W Typical 26,650 Coremark 45,330 Coremark Per core Per core SpecINT SpecINT Per core SpecFP Per core SpecFP

2x A53 1.0 GHz 2x A72 1.2 GHz 2.5 W Typical LS1023 LS1026 5.6 W Typical 8,360 Coremark 23mm x 23mm 15,000 Coremark Per core 780 pin Per core SpecINT FC-PBGA Package SpecINT Per core SpecFP Per core SpecFP

COMPANY PUBLIC 16 Detailed Example: LS1028A Processor With Dual Arm Cortex-A72 Core complex Security Core Complex • 2x 64-bit Cortex-A72 with Neon SIMD engine Security Engine • Speed up to 1300 MHz • Parity and ECC protected 48 KB L1 instruction and 32 KB L1 data cache ARM Trust Zone ARM ARM Cortex –A72 Cortex –A72 • 1 MB L2 cache with ECC protection Secure Boot Basic peripheral and Interconnect

16/32-bit DDR3L/4 • 2x USB 3.0 OTG controllers with integrated PHY Power Management 48 KB L1-I 48 KB L1-I Memory Controller • 2x eSDHC controllers supporting SD/SDIO 4.0 32 KB L1-D 32 KB L1-D • 2x CAN-FD controllers Standard interfaces • 8x UART serial ports 2x SD/SDIO/eMMC 1MB L2 - Cache Networking elements • Four Port TSN Ethernet Switch up to 2.5 Gbps on each port 3x SPI Coherent Interconnect (CCI-400) • Up to four SGMII supporting 1 Gbps

2x UART, 6x LPUART Accelerators and Memory High Speed interfaces Networking Elements • Up to one USXGMII supporting 2.5 Gbps Control 8x I2C, GPIO • Up to one QSGMII • Up to one RGMII 6x SAI, 2x CAN-FD 2.5 GbE 256KB SRAM SATA 3 • 2x PCI Express Gen 3 controllers FlexSPI PCIe 3.0 2.5 GbE • 1x SATA Gen 3.0 controller

8x Flex Timer Accelerators and Memory Control USB 3.0 2.5 GbE

TSN Switch TSN • 1x 16/32-bit DDR3L/4 Controller with ECC support up to 1.6 GT/s w/PHY Multimedia interfaces 2.5 GbE • Time Sensitive Networking (TSN) Ethernet Switch 3D GPU • Security Engine (SEC) 4K LCD Controller 2.5 TSN GbE PCIe 3.0 USB 3.0 • QorIQ Trust architecture: Secure boot, ARM Trust zone and security monitor w/PHY eDP/DP Phy 1 GbE Qualification • Commercial and extended temperature (support for 125C Tj) Package Target Applications: • 17x17mm, 0.75mm pitch FC-PBGA • Industrial Control, PLCs, Gateways • IoT Gateways Power • Automotive • Human Machine Interface • 5W TDP • Professional Audio/Video COMPANY PUBLIC 17 Trust Architecture

Hardware based security features to ease the development of trustworthy systems Manufacturing Secure All QorIQ SoCs support Trust Architecture Protection Boot 8 1 General General DDR Strong Secure Purpose Purpose Controller Processor Processor Partitioning Storage HV MMU HV MMU 7 2 Battery Back-up

Security Fuses Coherent Interconnect PreBoot Loader 6 Security Monitor IOMMU IOMMU Tamper 3 Key Internal BootROM Detection Protection Secure Debug Power Mgmt SEC Engine Mgmt Control Controller 5 4 SD/MMC Crypto, RNG Real Time Debug 2 QMan FMAN Tamper SPI DUART I C Watchpoint Secure Keys WRIOP Detect(s) IFC USB SATA BMan Perf CoreNet Secure Key Monitor Trace UID, Runtime Eth, Clocks/Reset Aurora Debug Revocation Integrity Check AIOP PCI CCSR GPIO Security Data-path sub-system sub-system

COMPANY PUBLIC 18 Generalized Comparison Between i.MX and Layerscape

Layerscape i.MX

• Scales to higher performance • Scales to lower cost/power • More PCI and Ethernet • More graphics and multimedia • Industrial qualifications • Commercial, industrial, and auto • ECC on all memories qualifications • • Linux, OpenWRT Linux, Android

COMPANY PUBLIC 19 Nexcom NSA 3640 For Midrange Edge Systems

• Processing and memory − NXP® OorIQ™ LS2088A, 8 Cores − Network accelerators − 5x DDR4-2400 DIMMs slots − Registered ECC-DIMM or U-DIMM • Interfaces − 4x 10GbE copper ports − 4x SFP+ ports − 1x PCIe Gen3 x8 or 2x PCIe Gen3 x4 slots − 2x SATA 3.0

COMPANY PUBLIC 20 Delta Networks LS104x Design for Edge Systems

• Applications • Networking Elements − Edge gateways − 1x GE/SFP Combo port for WAN(optional) − Security appliances − 1x GbE port for WAN − vCPE − 4x LAN GbE ports from 4p switch − 2 port PoE at/af(optional) • System − Desktop Fanless System − Layerscape LS1043A or LS1046A − On board 512MB DDR3L − On board 512MB SLC NAND Flash − 1x USB2.0, 1x USB3.0

COMPANY PUBLIC 21 Start Developing TSN on LS1021A-TSN Enhance with LS1028A

LS1021A-TSN LS1028A • TSN features • New TSN features • Time aware shaper (802.1Qbv) • Frame Pre-emption (802.1Qbu) • Per-stream filtering & policing (802.1Qci) • Frame Replication and Elimination (802.1CB) • Credit-based shaper (802.1Qav) • Cut-through Switching • Time synchronization (802.1AS) • Cyclic Queuing and Forwarding (802.1Qch) • 802.1AS-Rev • System features • 4 switched Gigabit Ethernet TSN interfaces • New system features • Arduino shield for IoT wireless integration • 4 CAN ports • Expandable IO – mini PCIe, SATA, USB 3.0, SD Card, GPIO Supported by one SDK – Open Industrial Linux

COMPANY PUBLIC 22 Intelligent Traffic Controller

• NXP is at the forefront of smart city/transportation development

• NXP has a V2X demo using multiple NXP products

• One Layerscape processor can fuse multiple inputs to control an intersection

COMPANY PUBLIC 23 NXP Edge-Box NXP development kit for Edge Computing

Customer Solution EdgeScale.org Device and Application Management App App App Cloud Application Development NXP SW Platform

Application Framework • NXP Edge-box • Add-on peripherals Linux Platform − 64bit 1-4 ARMv8 cores – up to 40K coremark. − Vector-engine card/stick Secure Firmware − 2-4 GE ports, 802.11ac/n Wifi, BLE. − Camera/audio − USB, SD card, M.2, UART, GPIO, mini-PCIE for − Wifi/LTE/BLE/LowPAN modules expansion/connectivity. − Pre-provisioned UID, secrets, OTA install/provisioning. • Software ecosystem − Multiple SKUs − Ready to run AWS GG, Azure, Aliyun, Google apps. ▪ Edge-box standard – LS1043/6, 200-400$ − EdgeScale cloud-based device and application ▪ Edge-box lite – LS1012, 50-100$ management. ▪ Edge-box display – LS1028, 100$-200$ − EdgeScale cloud-base application development. ▪ Edge-box extreme – LS2088, $500 − Choice of Ubuntu, Yocto or OpenWRT distros

COMPANY PUBLIC 24 Extreme Operating Conditions

10 Continuous operation at high Year temperature

• Product Life Application Notes • Extreme temperature conditions • -40o C cold start • 70-85o C ambient operating conditions • Up to 125o C junction temperature • Low power consumption for fanless designs • Small footprint for space-constrained designs

COMPANY PUBLIC 25 Supply Longevity

Industrial applications require product longevity • Long product lifecycles • Special product certification required

NXP Industrial Application Processors • 10 and 15 year supply longevity options • Formal program with products listed at www.nxp.com/productlongevity

COMPANY PUBLIC 26 NXP offers a broad portfolio of Layerscape processors Key Layerscape integrates functions to reduce Take-Aways system cost and power

Layerscape delivers class-leading performance

NXP has reference designs and ODM partners to accelerate customers’ TTM

Layerscape is robust and available for the long term

COMPANY PUBLIC 27 Layerscape Software

COMPANY PUBLIC 28 NXP Solution Expertise

Home SMB Router vCPE, NFV Edge WLAN-mesh Industrial Gateway Solution Solution Compute Solution IoT Solution Solution Openstack CloudSolution Device NXP orchestration Mgmt QoS, Protocol, TSN Linux Endpoint Solutions L4-L7 DPI stacks integration KVM, DPDK AWS, Azure Mgmt Fast-Path IPsec, SSL Virtual IO Secure Secure Security, offload offload offload Platform Backhaul Connectivity

Linux Customer Partner Open OpenWRT ONIE/ CentOS Yocto Distributions Linux Linux IL LEDE ONL Ubuntu

Open-source Kernel Uboot UEFI DPDK OpenSSL Others Software

NXP SDK Drivers - (Platform, Storage, PCIE, Ethernet, Crypto) Tools - (Trust, Manufacturing, Debug, Configuration)

NXP Platforms LS/LX/LA-series

COMPANY PUBLIC 29 Embedded Ecosystem is Evolving

Paradigm shift in embedded development Traditional – Multiple Open-Standards – Common - Traditional: Vertical, platform-specific solutions Closed Ecosystems Open Ecosystem - Take long time to market - Require large development teams Vendor-Specific Deployment OP-NFV, Open-Stack, ONIE - Are not re-usable across platforms

Proprietary NPU, Data-Path API Linux User Space, DPDK/ODP/VPP - Emerging: Standard, Open ecosystem

- Focus on applications

- Leverage software re-use Vendor BSPs, firmware Off-the-Shelf Distros - Less time spent on hardware

Software is in the driver’s seat Complex Data-Path Simple NIC Intelligent, Virtualized Network IO Applications are decoupling from hardware

Power MIPS ARM x86 ARM, x86 Accelerators

COMPANY PUBLIC 30 NXP’s Efforts to Bridge the Gap

• Driving ARM to port applications/libraries. Open-Standards – • Support OP-NFV, OpenStack, ONIE based orchestration. Common Open Ecosystem • Working with ARM and Cisco for vRouter. OP-NFV, Open-Stack, ONIE • Driving both ODP and DPDK. • Best in class acceleration and performance. • Support both VM and container-based virtualization. Linux user-space, DPDK/ODP/VPP • Added support for OpenWRT, ONL, Ubuntu, OpenIL. • Server-like UEFI bootloader. Off-the-shelf Distros • New SDK more distro-friendly.

• Significant push in upstreaming. • Virtualized DPAA2 ready for NFV/SDN, easy to use. Intelligent, Virtualized Network IO • AIOP software packages to ease adoption.

• Reuse standard ARM cores. ARM, x86 Accelerators • Leverage Linaro kernel, gcc work. • Comparable performance to Xeon-D and Atom. COMPANY PUBLIC 31 Common Software Ecosystem Between x86 & ARM

Open-Source Distro Repos Edge Apps Customer Management AppsSSL Ubuntu, Debian AWS GG Routing,Apps SDWAN Orchestration

VPN CentOS, Fedora Azure IoT Firewall, DPI Cloud frameworks – AWS, Azure, Aliyun DPI/Firewall OpenWRT Aliyun VPN, SSL Customer Customer solution KVM DPDK, ODP OpenSSL Provisioning Apps Customer SW Docker VirtIO, VFIO Linux NW Stack I/O NXP Secure Device, App Mgmt Services OpenSource SW

Virtualize Ceph Upstream drivers StrongSwan Networking Intel NXP OP-NFV, OpenStack HW HW Boot/OS U-boot, UEFI, ONIE CentOS, Ubuntu, Debian Yocto,OpenWRT Customer Boards

Silicon LS-series – 1 to 16 ARMv8 cores Denverton, Xeon-D series – 4 to 16 cores

Hardware RDB Blades, iNICs Servers

COMPANY PUBLIC 32 NXP Linux Software

rd NXP Turnkey Solutions 3 Party Solutions Solutions (Security, ODM designs, Wifi, Customer Solutions (Gateway, Secure Router) Switch)

Layerscape Custome Partner Yocto OpenWRT ONIE/ CentOS Linux SDK r Linux Linux Project LEDE ONL Ubuntu Integrations

Custom Linux Standard Linux Distros

Open-source Kernel Uboot UEFI DPDK OpenSSL Others Communities (linaro, kvm, (kernel.org) (denx.de) (tianocore) (dpdk.org) (openssl.org) docker, OPNFV) NXP drivers NXP tools NXP (Platform, Storage, PCIE, Ethernet, Crypto) (Trust, Manufacturing, Debug, Configuration) Components

COMPANY PUBLIC 33 Different Ways to Get Linux LS components Layerscape flexbuild User-space Components Build-systems 1 SDK Ubuntu user-space Ubuntu Platform Support flexbuild packages Boot-loaders components Roll-your- Kernel.org LTS 2 Choice of build-system kernels own Linux Customer Choice of user-space NXP drivers Customer Applications build-system User-space libraries yoctoproject.org Openwrt.org components LS components Tools OpenWRT Yocto Buildroot OpenWRT Configuration (buildroot) Yocto user-space packages OpenWRT Centos.org OpenIL Ubuntu Yocto 3 (Open Ubuntu Yocto LS components LS components Embedded) packages Standard CentOS buildroot Linux CentOS user-space LEDE + Industrial Distros ONL Canonical Kernel.org components LS components 4 Open-source community Debian Ubuntu Denx.de ONL user-space Ubuntu user-space

COMPANY PUBLIC 34 Platform Support Lifecycle Kernel.org • Prior to upstreaming, NXP will release provide standard SDK support.

Submit New NXP • NXP is actively committed to platform platform – Add bugfixes to upstreaming. drivers in SDK kernel.org • Once upstreamed, NXP recommends to get support via community.

• Premium Support and Frozen Upstream new Platform now Branch options a choice drivers to part of throughout via NXP kernel.org kernel.org Commercial Services

COMPANY PUBLIC 35 Layerscape SDK

Ubuntu public repositories • Components Even more packages − Freely available from public repositories, no need to download an ISO Ubuntu user-land Layerscape − Modularized with discrete boot loaders, user- tools space libraries, tools, config User-space applications Target Platform apt- build configuration − Clean layering, with separate patches identified get User-space libraries env User-space by platform/IP on top of open-source drivers − Git updates of periodic releases and interim updates LTS Kernel − Support for 2 recent LTS kernels Linux Network Stack File-system Visualization Power Mgmt Upstreamed Layerscape support • Living with the SDK − Multiple boot mechanisms with recovery via SD Ethernet Crypto PCI, USB, SATA, Flash Mature platforms − Run-time upgrades with apt-get or build from source Layerscape kernel patches − Latest and greatest Layerscape tools and drivers New drivers New platforms • LSDK is not a Linux distribution but a Board-specific reference integration with a Ubuntu user land Bootloader – Uboot, UEFI Init-time configuration Other tools

COMPANY PUBLIC 36 OpenIL.org G TSN A IEEE1588 Hard Real-Time Industrial-Grade Security F Open Industrial B Open Software Repository and Community

Industrial Networking E Protocols C HMI Applications D Manageable via industrial control and networking protocols COMPANY PUBLIC 37 Edge Computing Ecosystem Edge computing is the Customer Solution application of cloud technology AWS Azure Aliyun outside a large data center. App App App Cloud NXP SW Platform Network Application Framework Linux Platform Secure Firmware Smart Nodes NXP i.MX (Machine-learning) Smart-nodes can run targeted Edge applications Network Infrastructure Edge Gateway NXP Layerscape NXP Layerscape (Aggregation, Analytics) Data Nodes (Control, Analytics, Machine-Learning) Edge applications can also NXP LPC/Kinetis run on access – e.g., base Families Gateways are a natural host for edge stations, central office – computing – right balance of compute, costlier pipe, higher latency. connectivity and storage COMPANY PUBLIC 38 Security Is Multifaceted PRIVACY preventing eavesdropping, data exfiltration EXAMPLE Encryption

AUTHENTICATION guaranteeing identity EXAMPLES User IDs (e.g., passwords) Hardware IDs

ACCESS CONTROL limiting use and communication EXAMPLE Access control lists

SYSTEM INTEGRITY AND AVAILABILITY ensuring data and code accuracy preventing service theft and denial, cloning EXAMPLES Secure boot Runtime integrity checking

COMPANY PUBLIC 39 Examples of IoT Hacks

Automobile Point of Sale Home Routers

• Telematics hacked to: • Memory scraper • Root access and

− Eavesdrop installed on POS program installation

− Control ECU • POS exfiltrates credit • DDOS launched card info

Vulnerabilities Exploited

Flawed Authentication Promiscuous Communication Stack Overflow No End-to-End Security Unsigned Code No Memory Protection No ACL

COMPANY PUBLIC 40 Management and Security Challenges Solution: Cloud-based Management & security for edge • Manage devices, apps remotely • Secure provisioning, upgrades

Traditional PC, mobile devices • Multiple authentication mechanisms • Cloud based security and application management

Edge computing devices • Traditionally embedded devices • Not physically accessible, or lack display • Can be many (10s, 100s, 1000s) per user

COMPANY PUBLIC 41 Security Requirements for the Edge 05 Secure Chain of Trust Application Deployment • Security starts with Secure 04 Container hardware root of trust Deployment • End-to-end security is a chain of inter-locked 03 Secure Device security elements Monitoring Device Management 02 Secure • Secure manufacturing Enrollment • Secure enrollment 01 • Secure device monitoring and Secure firmware management Manufacturing • Secure container deployment Credentials may be installed in on- • Secure app management and chip Layerscape fused memory or deployment via external Secure Element.

COMPANY PUBLIC 42 EdgeScale for Device Management

Cloud Applications EdgeCloud Applications Applications Customer have choice for Application Management Embedded Edge Compute Applications Framework Application − AWS, Azure, Aliyun, Management Google Service − Home-grown or 3rd Party

Operating System Applications

EdgeScale provides

Device Management Device − Device Management Mgmt Management − Security via Hardware Device Provisioning Service Root of Trust

Device Edge Gateway Cloud

COMPANY PUBLIC 43 EdgeScale – Flexible Architecture

Customer Private Multiple compute Cloud and usage models Edge Compute Frameworks

Dashboard, CLI RESTful APIs

Common device management service EdgeScale Device Management Services

SDK - EdgeScale SDK - EdgeScale 3rd Party SDK Hardware Root of Trust Trust Architecture Trust Architecture Secure Element Multiple platforms Other SoC supported Layerscape i.MX vendors

COMPANY PUBLIC 44 EdgeScale Architecture

Device Cloud

Traditional Edge Computing Application Management Services Embedded Azure Cloud Azure Azure AWS Apps Ali Apps DMS AWS Apps AWSApps Apps AWS Cloud AWS Apps Embedded AWSAWS Apps Apps AWS Apps AWS Apps AWS AppsApps Portal Applications AWS Apps AWS GG Alibaba Azure AWSAWS Apps Apps Alibaba Cloud DMS Portal User-space, Docker, Kubernetes DMS Ali Apps AWS Apps Portal AWS Apps Linux

Device Management Agent Device Management Services Secure Firmware Container Deployment Service Dashboard Firmware Build Services Secure End-User Factory OTA Secure Firmware Service Database Manufacturing Provisioning Provisioning Agent Device Secure Enrollment Service Database Code-signing Tool Firmware Secure Device Management Secure Device Monitor Service Database

NXP Provides Open-Source Embedded Apps Cloud Apps COMPANY PUBLIC 45 How NXP Makes Edge-Framework Security Simpler

Today - Manual Framework Deployment EdgeScale automates frameworks’ key Generate key pair at cloud distribution and software deployment Get key pair and Cloud device image company

SSL talk Cloud Copy key pair Serverless key/cert to device EdgeScale functions service

Copy image ManageManage Lambda serverless to device Provide signed Container functionsfunctions on GG in NXP devices binary containercontainer Provision certificate Validate & launch Container binary EdgeScale solution: IOT Edge SDK Framework • Automates edge framework software and EdgeScale Agent key distribution Secure Firmware • Deploys framework core as signed Docker image • Provides on-device trust computing and secure storage for apps

COMPANY PUBLIC 46 NXP Works With Cloud Companies’ Edge Frameworks

• Amazon Web Services (AWS) − Greengrass integrated with Layerscape − Edge demos using Greengrass and AWS − EdgeScale integrated with AWS and GG − http://media.nxp.com/phoenix.zhtml?c=254228&p=irol- newsArticle&ID=2289486

• Microsoft Azure IoT Edge − Azure IoT Edge supported on Layerscape − Focus on secure execution and Docker − http://media.nxp.com/phoenix.zhtml?c=254228&p=irol- newsArticle&ID=2334845

• Alibaba Cloud (Aliyun) − NXP processors uniquely support Alibaba TEE OS − Alibaba and NXP plan to jointly develop smart manufacturing and smart city applications − http://media.nxp.com/phoenix.zhtml?c=254228&p=RssLanding&cat =news&id=2322324

COMPANY PUBLIC 47 NXP understands and addresses key trends Embedded systems Key IoT and edge computing Take-Aways NXP understands and addresses security and device-management challenges NXP Trust Architecture EdgeScale

NXP leads in collaborating with cloud companies on edge frameworks Accelerates NXP customers’ time to market Improves security

COMPANY PUBLIC 48 Edge Computing Applications

COMPANY PUBLIC 49 Building Automation

• Killer app is linking systems via an edge-computing nexus • HVAC + surveillance − Start HVAC when employee arrives • HVAC + weekly weather forecast − Plan vs react • Ingress + egress security cameras − Did someone leave behind a backpack? • Sensors + fire alarm − Warn if unusual electric load, items piled up near boilers, etc.

Photo by Franklin Heijnen © 2015 https://flic.kr/p/uf7BkK COMPANY PUBLIC 50 Licensed per CC BY-SA 2.0 https://creativecommons.org/licenses/by-sa/2.0/ Home Automation • Fire/burglary alarm • Remote unlock • Security cameras • Remote doorbell • Thermostats • Smart speakers • How edge computing differs − Learning done locally − Local intersystem coordination − Shared learning (e.g., voice, face recog)

Photo by “m01229” © 2015 https://flic.kr/p/mK7qJd − Automation enabled when house Licensed per CC BY 2.0 https://creativecommons.org/licenses/by/2.0/ offline − Privacy

COMPANY PUBLIC 51 Warehouse Automation

Robots shuttle shelves

Conveyer belts transport goods

Machines dispense tape and boxes

Edge computing orchestrates everything

COMPANY PUBLIC 52 Medical and Healthcare Monitoring and Automation

• Surgery robots – Parallels factory automation development – Edge nodes monitor and analyze sensors – Edge nodes coordinate robots • Patient monitoring – Edge computing enables offline analysis – Edge computing enables sensor fusion

COMPANY PUBLIC 53 NXP is Bringing EdgeScale to an Alligator Farm

• Alligator farm has pole- or tower- mounted observation hardware

• Today: firmware updates requires dangling from a helicopter with a PC and cable

• Same constraints apply to industries with remote equipment like oil/gas

• EdgeScale remote management enables low-cost remote updates

COMPANY PUBLIC 54 NXP Has Shown Enterprise and Retail Video Analytics

• Customer check-in −Face and ID correlation using AI

• Retail analytics using AI −Hot spot detection −Face recognition −Correlate items browsed with shoppers

−Shoplifting detection Source: NXP

COMPANY PUBLIC 55 NXP MCUs Can Instrument Factory Equipment

• Predictive maintenance • Condition monitoring for anomalies − Vibration, temp, sound, etc. • Chaotic outer control loops − Neural nets to predict chaotic systems • AI-based visual inspection • Machine control using sensor data • Edge computing advantages − Distributed preprocessing ▪ Reduce data, spread computational load ▪ Low latency Source: NXP

COMPANY PUBLIC 56 NXP Is Already Engaged in V2X Development

• Aggregate data − Sent by cars − Observed by fixed cameras − Observed via radar − Sent by other V2X nodes

• Analyze and control objects − Track objects − Predict objects’ behavior − Control behavior with signaling − Manage pedestrian and vehicle queues

• Identify carpool and other violations

COMPANY PUBLIC 57 How Can Edge Computing Be Used in Aircraft?

Real-time analysis of engines and other flight data? • Preventative maintenance • Fuel-economy optimization

In-cabin entertainment and services? • Management of seatback systems • Security/isolation of systems • Health and safety Edge computing is broadly applicable

Key AI is a killer app for edge computing Take-Aways NXP is actively developing edge applications for demonstration and with customers

COMPANY PUBLIC 59 Conclusion

COMPANY PUBLIC 60 How Might Edge Computing Transform the World?

• Bigger role for software in traditionally hard industrial systems − Affects what industrial companies offer (e.g., GE Predix)

• Software development becomes more IT-like − Abstract − Leveraging frameworks, containers/VMs − Decouple from hardware − Easier to manage

• Bigger role for cloud service providers

• Improved cost, performance, and safety by fusing multiple systems’ inputs

COMPANY PUBLIC 61 Machine-generated models factor in more data than man-made ones How Might AI Better models yield better analytics, Transform the revealing hitherto hidden insights New, gee-whiz features (e.g., remote World? unlocking of your house) Reduced cost (e.g., predictive maintenance) Improved safety (e.g., stopping worker without hardhat from entering) Bigger role for software in traditionally hard industrial systems – Continues the trend that started with replacing analog with digital control

COMPANY PUBLIC 62 How Might Processors Change?

• Cross-vendor compatibility is increasing owing to: −Arm −Containers/VMs −Frameworks

• Accelerators will become transparent to software

• AI accelerators will become common even in small-scale processors

COMPANY PUBLIC 63 Layerscape: a broad portfolio of Arm- Summary: compatible processors integrating I/O NXP Linux and enabling software Layerscape ODM and reference designs Offerings EdgeScale for edge/IoT node management

Cloud framework support

Demos and early-stage success in edge computing

COMPANY PUBLIC 64 Transform Your Business with Edge Computing and AI Using Layerscape!

COMPANY PUBLIC 65 www.nxp.com NXP, the NXP logo, and NXP secure connections for a smarter world are trademarks of NXP B.V. All other product or service names are the property of their respective owners. © 2018 NXP B.V.