Iot gateways march 2015

MARCH 2015|

IOT GATEWAY CHALLENGES AND PLATFORM SOLUTIONS

SRINI GADGIL SENIOR SYSTEMS ARCHITECT

MARCH 2015

2 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

IOT APPLICATIONS

Enterprise/Building Management

Smart Cities Smart Home

Health Care

Smart Grid Industrial Automation

2012 ( Source Cisco)

8.7B

Connected Devices

2020 (Gartner)

26B

3 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

IoT NETWORK WITHOUT GATEWAYS OR “LIGHT GATEWAYS” Cloud Mobile / Wireline Access Network

Data Center Things

Control

Devices

Cellular

Network

WAN

Network

IOT PAN/LAN

Network

4 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

ISSUES WITH NON-GATEWAY NETWORK / CLOUD CENTRIC APPROACH

Network issues

‒ Cellular Network Limitations : Range, Bandwidth, Reliability in Buildings, power, penetration

‒ Cost of Spectrum

‒ Capacity

‒ Cost

No Gateway / “Light” non-processing Gateway

‒ No redundancy if the Network Connection/Cloud App. is lost

‒ No local processing : this will impact processing time

‒ It will be difficult to implement uniform security policies.

5 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

ENTERPRISE SOLUTION FOR IOT WITH GATEWAYS

Security + Access Control

Business Premise IOT Network

Cloud

Data Center

HVAC + Power + Lighting

Irrigation and Water Mgmt.

Control Devices

IOT Access Gateway

Solar Energy Management

IOT Gateway

• Local Intelligence : provides redundancy during network/cloud outages

• Faster local processing

• Future proofed for enhanced Applications Support

• Mitigates access bottlenecks

• Uniform security policies can implemented

Cellular

Network

WAN

Network

Mobile / Wireline Access Network

6 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

IoT GATEWAY CHALLENGES

IOT Access Gateway

Multi-protocol Connectivity

Support all the protocols used to connect sensors, actuators, control devices and

network uplinks

High Performance Processing

For certain specialized workloads such as Machine Vision and Biometric processing for

fast response time

Manageability

Support SDN enabled Networking stacks for configuration, management, automation,

orchestration.

Cost effective solution

Gateway is cost optimized for lower BOM cost. Lower power, fan less operation.

CPU throughput / redundancy

CPU throughput to support standalone operation in case of loss of connection to the

cloud

Scalability

Platform HW and SW architecture must be built for addition of future applications and

related devices

OS Platform

OS platform should have wide acceptability and must support a diverse device and

applications ecosystem

Device Ecosystem

Gateway platform should support rich ecosystem of present and future devices :

sensors, actuators, control devices

Local Processing vs Cloud

The Gateway Architecture should partition this very carefully the main tradeoff is

between faster response time vs cloud/big data capabilities

7 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

AMD IOT GATEWAY ARCHITECTURE

I/O : USB, Ethernet, UART, PCIe, RS-485, etc.

Drivers

Sensors

Actuators

Devices

Monitoring And Control

Short / Medium Range PAN/LAN Access Network

WiFi, DASH7, ISA100

ZigBee, 802.15.4

Bluetooth

Ethernet, EtherCAT, Modbus, Profinet

Device Drivers

GPU based image processing

Wired/Wireless Protocol Stacks

Protocol Conversion

Image Recognition

IOT Applications

Data Acquisition and Logging

Access Gateway Software Functions

Drivers

Ethernet /

MAN / WAN Networks

Drivers

Cloud Datacenter

LINUX / RTOS

Appliances

Wireless MBus

Wi-fi 802.11ax

Cellular

Satellite

T1/E1

Storage

8 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

AMD IoT GATEWAY SOLUTION BASED ON G-SERIES CPUS

IOT Access Gateway

Multi-protocol Connectivity

Supports all the protocols used to connect sensors, actuators, control devices and

network uplinks

GPU processing GPU processing can be

used for local fast response Biometric / Optical sensor image processing.

Supports HD displays.

Innovative solution for SDN, NFV

Supports SDN enabled Networking stacks. Some NFV functions such as Security

implemented on GPUs

Cost effective solution

CPU + GPU solution is cost optimized for lower BOM cost. Lower power for fan less operation.

IOT Gateway

SPI EEPROM

PCIe

DDR3 DIMM

Enet MAC/PHY

Module

Module

Module

USB

PCIe/USB

USB

USB

USB

CPU throughput

CPU throughput will support standalone operation in case of loss of connection to the

cloud

Scalability

Platform HW and SW architecture is built for addition of future applications

9 | AMD EMBEDDED SYSTEMS | MARCH 2015 |

AMD IOT GATEWAY SOFTWARE STACK

AMD HSA Processor

OS User Space

CPU bound IOT Functions

GPU

GPU (Acceleration) bound IOT Functions

CPU

Shared Memory

MAC/PHY

Applications

Reporting Monitoring Biometric Image Processing

Module Event Notification

Device Connectivity Drivers

OS Kernel Space

Power and Lighting Control

Access Control

And Security

Communication Protocol Stacks

Data Acquisition and logging

Power Metering

Cloud Application for power

Metering

Cloud Application for

Security Applications

Cloud Applications and Databases

Power and Lighting Control

Power Analytics

Access Control

And Security

Mobile App