Data in Motion - tech-intro-for-paris-hackathon
-
Upload
cisco-devnet -
Category
Technology
-
view
515 -
download
4
Transcript of Data in Motion - tech-intro-for-paris-hackathon
Thierry Gruszka
Senior Technology Manager
4th Nov. 2015
Workshop Cisco DevNet Hackathon
Data in Motion - DMo
DATA !?
Wisdow
Knowledge
Information
Data
• Je ferais bien de m’arrêter Control
• Je conduis et le feu tricolorevers lequel je me dirige passeau rouge
Context
• Le feu tricolore à l’Angle sud de la rue Tom et de l’avenue Jerry vient de passer au Rouge
Meaning
• Rouge, 192.234.235.245.678, v2.0Raw
DMo
© 2015 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 3
• Data in Motion is an IoT software product that runs in the network to transform raw data from sensors and endpoints into actionable information.
• Data in Motion enables to build scalable IoT solutions
Data in Motion Overview
Cisco Confidential 4© 2013-2014 Cisco and/or its affiliates. All rights reserved.
Data in Motion at the Edge
Input Data
Store raw data or filtered data for general data management
Analytics
Cloud and Data Centers
Generate Actionable Events and learn new rules
Cache raw data or abstracted information
(e.g. indexed data)
Data in Motion:
Analyze First,
Optional Store
Input Data
<XML>
Rules can express:
Predicates and Filters
Data / Information
conversion
Summarization
Pattern Matching
Categorization &
Classification
Event Trigger analysis
Notifications
</XML>
sensor Router/Switch
Traditional Data
Management:
Store First,
Analyze later
Cisco Confidential 7© 2013-2014 Cisco and/or its affiliates. All rights reserved.
Mining
+ +
• Data reduction and summarization
• Event triggered Analysis
• Edge data subscription model
• Predicates
• Policy driven
• Categorization and classification (indexed)
• Content re-purposing
• Data understanding at the edge
• Programmability at the edge
• Connectivity
• Multiprotocol
• micro-CDN (store & forward)
1 2
VEHICLE WEIGHT: 08 TONS
GROSS WEIGHT: 16 TONS
Customer: Anglo American
Use case: Track truck pressure tires for load monitoring
Targeted Platform: 819H
Software Equipped: Data in Motion
Release Date: November 2013
Cisco Confidential 8© 2013-2014 Cisco and/or its affiliates. All rights reserved.
Smart Agriculture
HUMIDITY: 40%
TEMPERATURE: 82F
ACTION: SPRINKLER
ACTION: SPRINKLER
ACTION: SPRINKLER
• Content re-purposing
• Data understanding at the edge
• Programmability at the edge
• Connectivity
• Multiprotocol
• micro-CDN (store & forward)
Customer: University Space Research Association (USRA) for USAID
Use case: Frost Detection for Crop Management in Third World USAID Programs
Targeted Platform: UCS-E/C and CGR 1K
Software Equipped: Data in Motion
Release Date: April 2014
1
Cisco Confidential 9© 2013-2014 Cisco and/or its affiliates. All rights reserved.
Monitoring
Actual data is sent only
when system is at fault
Event is detected right
at the edge
EVENT: LEAKAGECONTAINER 107
Pressure : 2psi
Humidity: 14%
Temperature: 35F
Use Case with Event Notification (Surveillance)
Supporting various data Sources:
webcams, files with Data in
Motion.
Two major search capabilities
Searching people or objects
example: Search people carrying
a backpack and having short hair.
Searching scenes
example: Two people carrying
backpack within the same view
of a camera. One of them is
wearing black shirt and the other
is wearing white shirt.
Train jubatus with annotated training
data set
Data in Motion
…
Automatically add tags
using Machine Learning.
Search tags with temporal
Information. Full text search
Is also supported.
video analysis system
Jubatus learns which tags to set
for each person or object.
All you have to do is to provide
annotated data.
This system allows users to search
people or objects in their video
flexibly by using Machine Learning
and a search engine.
Example Use-case with video
• Purpose
• Annotate people’s appearance and behaviors
• Detect anomalies and make search index
• Application
• Alarm for crimes and suspicious behaviors
• Help investigating criminals on the run
• Search and locate suspects by characteristics
• Advantage
• No need to monitoring by human eye
• Instant search by characteristics tags
• No need to check all videos for massive hours
• Purpose
• Annotate customers’ appearance and behaviors
• Estimate their profile and intention in detail
• Application
• Detect unseen demands to serve
• Analyze POS data with detailed categorization
• Optimize items, layout and shopping process
• Advantage
• More precise and dynamic than analyzing only
POS and membership information
(1) Surveillance (2) In-store behavior analysis
Data in Motion Data Sheet
Data in Motion plane
Data (Packets)
Data Acquisition & Transformation
Information
Rules/Patterns
Data to Information Capabilities• Event Detection & Aggregation
• Rule-Based Data Normalization
• Dynamic Sensors Polling
• Unstructured Data Understanding
• Data & Information Caching
• μ-CDN (Controlled Distribution)
• Pub-Sub API (Eclipse IDE)
Supported Platforms• UCS-E/Blade
• CGR-1K
• C8xx with Iox Packaging
Use Cases• Data Reduction and
Compression
• Sensor Virtualization and
Plug & Play
• The API interfaces with the user's programing environment. The user writes a software program that specifies what data s/he is interested in.
• The API helps the user translate rules in open standard JSON format encapsulated as a REST message that can be understood by the API.
• A key part is the format of the JSON messages used to express a rule. The API to the edge device of interest using a RESTfulcommunication paradigm then sends this rule. This is the main publish part.
How does it works…
Data in Motion is a native application in Cisco IOx
IOS +IOx SDK
Virtual Machine
Linux OSData in Motion+IOx
Application
Management
Control Plane Data Plane
Data in Motion Policy / RulesA true Real time transaction with a Model Definition
• Dynamic Data Definition involve the relationship of three simple concepts
• Pattern Extraction real time content indexing
• Condition Rule Engine to query over index & algebraically
• Action Many, including data transformation and engaging network
connectivity
• Ultimately this breaks down into data understanding and of:
D3
Meta (1)
D3_Id, Context_ID, Processing Method (Timer, Cache)
Network (01)
Filterby: (protocol {tcp/ip, UDP}
Source/Dest IP, Source/Dest Port (multiple ANDed)
Decode: (variable A=first 8 Bits, var B=next 16 bits, etc….)
Application (01)
Filterby:
Protocol: http
Field: content-type:json, etc.
Content
Example: variable Temperature>56
Action (>1)
Type: Primitive
payload
Header
Type: Procedure
FetchData
Gpsupdate()
syslog
Type: Timed
FetchData
Gpsupdate()
syslog
• Network Meta Data • Application • Content
• Action(s)