It is often said that if the GPS was invented today, the company that would’ve done it would also invent and provide Waze.
Nowadays more and more device vendors realize that instead of providing a commodity device, it is much more valuable to provide a long term service around it. By converting a basic lighting bulb business to a lucrative smart lighting experience Philips hue marked the way, and everyone followed.
But how to do it? Choose one of the 300+ IoT platforms out there? Go for DIY? And what’s the take of the open source and cloud providers in this puzzle?
Here is a guide to the perplexed.
2. Large scale SW is like Condensed Matter
Million-scale users, 10000s devices, 1000s TPS
◦ AWS partner
IoT clouds maker for device makers
◦ Fully managed service
◦ Enables device makers to focus on their core
Customers - global $Bn companies to start-ups
About Me and Softimize
7. Smart home as example
For the user
◦ Centralized monitoring and control
◦ Buy the GW/Apple TV
For the device maker
◦ SDK for device drivers – Groovy
◦ Direct or cloud 2 cloud
Great for motion sensor, door lock,
temperature sensor, On/Off switch,…
Verticals
8. Generic
For the user
◦ Manual control – DO recipes
◦ Automated control – IF recipes
◦ >200 integrations to SaaS
◦ No status views
For the device maker
◦ Cloud 2 cloud integration
Great for any-integration
9. User-centric
Centralized operation
• Commoditization
• Lowest common denominator
• Launching new features – not there
• No stream processing
• Lack of device management
• Device configuration
• OTA FW
• No usage analytics
12. Bottom-up
Most of the 8 capabilities
◦ Entity builder
◦ Telit – data consumption monitoring
Few cloud – cloud integrations
On top of AWS (Ayla, Telit)
10s of customers
For a device maker
◦ Require using their embedded board
◦ Create your own web/mobile apps
◦ Device agent, mobile libraries
◦ Cloud side - REST API
13. IoT application generator. Web/mobile.
◦ UI to create entities, queries, pages with
fields.
◦ Little real programming, though possible to
add JARs
For a device maker
◦ Send info with ThingWorx API, or us
◦ “ThingWorx-ready” HW components
◦ Upgrade FW with Axeda
Top-down
14. Time to market and dev cost – reduced by up to 50%
Stability of a platform used by many others
Effective to create an organization’s “Smart Things”
• Licensing >$10 per device per year
• With over 300 PaaS, What if you were wrong?
• Unfit database, deployment location, API
• Data leakage
• Features limit
• Dead (Parse.com)
16. AWS IoT – MQTT/HTTP GW, Device SDKs (C, JS,
Arduino), Rules engine, Device shadow & registry
IoT Suite
Watson IoT – MQTT GW
Will soon be there
DIY - Device Connectivity Cloud Services
+ non-IoT cloud services
17. IoT
◦ Hobson, Kaa – IoT platforms
◦ Vorto (Eclipse) – Things modeler and code
generator
Generic
◦ Backend - NoSQL, Spark Streaming, Complex Event
Processing
◦ Web Client - AngularJS, React, Auerelia
DIY – Open Source Tools
18. User experience perfectly matches your vision
Low OPEX
Our Device
Heating
Cooling
Electronics
Mechanics
Hydraulics
Carbonation
Secret1
Secret2
UI
Design
IoT Service
IoT
SaaS
Web/Mobile Apps
Cloud
Security & Privacy
Billing & payment
SW Productization
Smart Home, Smart Health
User & Product Analytics
Social Networks
Our Core Skills,
Differentiation,
and Focus
New Skills,
No Differentiation,
Defocus
19. PaaS Platform In-house R&D
Reasonable time-to-market
Low investment
High OPEX
Vendor Lock
Long time-to-market
High investment
Reasonable OPEX
Exact Fit
Why Suffer?? Enjoy The Best of Both Worlds
20. TBBTM - Cloud-optimized Open IoT Platform
Avoid Lock
TBB is a private open source
It’s your IP, deployed on any
infrastructure and location
Minimize time to market
Out-of the-box features
Minimize OPEX - Down to 1/10!
Pay single-digit cents per sensor per month
No royalties to anybody
Minimal DevOps
Minimize investment
Pay only for your unique business logic
21. Sensor Hub
E
L
B
API
Server
K
i
n
e
s
i
s
Everything
DynamoDB
S3 or EFS
Session Files
Cloud Front
SQS
Agg
Trigger
Algo Agg
SQS
Agg
Ready
S3
Agg Out
Hub Logs
S3
Agg In
Auto Scaling
Auto Scaling
Auto Scaling
S3
DB
Backups
Redshift
Any SQL
BI Tool
Session Cache
RT Algo
*H1/2016 - MQTT (including AWS IoT)
TBBTM Data Streaming Implementation over AWS
22. Security Multi Tenant DB
Data Streaming Vendor Services
Management
TBBTM = Things Building Blocks
Push notificationsDevice Interaction
Access Control – laas, SaaS
Vendor and cloud provider protection
Site management – Multi device
Licensing – per Tenant. Trial license
Bulk versioned FW updates
Complex event processing
Real-time, sub second latency
Users | Devices and hierarchies
Encryption, Tenant isolation
Audit, Dashboard
Analytics – Failures, Usage patterns
Prediction – Churn, Upsell
Discover & Config – w/o wifi | Real time streaming | FW update
Security – Encrypt, Auth | Reduce energy & bandwidth
On Premise
MQTT, HTTP
Built for Clouds
Multi Cloud
Abstraction Layers for managed services
NOT using DevOps-hungry open sources
23. From Open Source to Your Product’s IoT Service
IaaS
PaaS
SaaS
Connectivity
Softimize TBB
Your IoT Service
Services
developed by Softimize
24. A Software Product by Softimize
Product Design
User types
Use cases
PRDs, Wireframes, Mockups
Production environment
Automated monitoring
Security hardening
Development
Cutting edge technologies over TBB
Performance under load (OPEX)
Quality and robustness
Maintenance
Fully Managed Service (24/7)
Cost control
Up to date stack
Pre-sale support
Product Release
25. Now You Can Focus on Your Core
Your R&D increases
your true IP
Softimize reduces
risk, TTM, and OPEX
29. Shopping List Architecture
Sensor
Hub
Raspberry PI
TBB RPI JS SDK
AWS
IOT
Kinesis
DynamoDB
Auto Scaling
TBB
Streamer
Sensor
Actuator
Actuator
E
L
B
TBB
Device
Mgmt
Auto Scaling
Web
REST API Client
TBB AngularJS SDK
MQTT,
Hub Certificate
Provision Things
Set/Get shadow
Push notifications
MQTTWebSocket
IoT Service app
Device management
31. Challenges with AWS IoT
Device Modeling - Hierarchies
◦ 201X – most edge sensors have no IP address. (BLE, ZigBee rule)
◦ Need to model a GW and devices under it.
◦ Workaround: flat model + certificate sharing
Device Modeling - Configuration and Status
◦ 2 sources of truth vs 1 Desired at Shadow– The device and the
cloud
◦ Solution – use Reported for status. Partial updates are available
Device SDK
◦ Exception when changing the shadow while
disconnected