The First Mile - Edge and IoT Data Collection With Apache Nifi and MiniFi

© Hortonworks Inc. 2011–2018. All rights reserved1
The First Mile – Edge and IoT Data Collection
with Apache NiFi and MiNiFi
Andy LoPresto | @yolopey
Sr. Member of Technical Staff at Hortonworks, Apache NiFi PMC & Committer
16 October 2018 Dataworks Summit Tokyo

Gauging Audience Familiarity with NiFi
“What’s a NeeFee?”
No experience with dataflow
No experience with NiFi
“I can pick this up pretty quickly”
Some experience with dataflow
Some experience with NiFi
“I refactored the Ambari
integration endpoint to allow
for mutual authentication
TLS during my coffee break”
Forgotten more about NiFi
than most of us will ever
know

Agenda
• What is dataflow and what are the challenges?
• Apache NiFi
• IoT Challenges
• Apache MiNiFi
• Exploration
• Community
• All slides provided online, so no need to transcribe

What Is Dataflow?

What Is Dataflow?
• Moving some content from A to B
• Content could be any bytes
• Logs
• HTTP
• XML
• CSV
• Images
• Video
• Telemetry
Producers A.K.A
Things
Anything
AND
Everything
Internet!
Consumers
• User
• Storage
• System
• …More Things

Moving Data Effectively Is Hard
“Data Pipeline” https://xkcd.com/2054/

• Standards
• Formats
• Protocols
• Veracity
• Validity
• Schemas
• Partitioning/Bun
dling
Data
Dataflow Challenges in 3 Categories
Infrastructure
• “Exactly Once”
Delivery
• Ensuring
Security
• Overcoming
Security
• Credential
Management
• Network
People
• Compliance
• “That
[person|team|g
roup]”
• Consumers
Change
• Requirements
Change
• “Exactly Once”
Delivery

Raise your hand if you want to maintain Python scripts for the rest of your life
Let’s Connect Lots of As to Bs to As to Cs to Bs to Δs to Cs to ϕs

What Is Apache NiFi?

• Guaranteed delivery
• Data buffering
• Backpressure
• Pressure release
• Prioritized queuing
• Flow specific QoS
• Latency vs. throughput
• Loss tolerance
Key Features
Apache NiFi
• Data provenance
• Supports push and pull models
• Recovery/recording
a rolling log of fine-grained history
• Visual command and control
• Flow templates
• Pluggable, multi-tenant security
• Designed for extension
• Clustering

Flowfiles Are Like HTTP Data
HTTPData FlowFile
HTTP/1.1 200 OK
Date: Sun, 10 Oct 2010 23:26:07 GMT
Server: Apache/2.2.8 (CentOS) OpenSSL/0.9.8g
Last-Modified: Sun, 26 Sep 2010 22:04:35 GMT
ETag: "45b6-834-49130cc1182c0"
Accept-Ranges: bytes
Content-Length: 13
Connection: close
Content-Type: text/html
Hello world!
Standard FlowFile Attributes
Key: 'entryDate’ Value: 'Fri Jun 17 17:15:04 EDT 2016'
Key: 'lineageStartDate’ Value: 'Fri Jun 17 17:15:04 EDT 2016'
Key: 'fileSize’ Value: '23609'
FlowFile Attribute Map Content
Key: 'filename’ Value: '15650246997242'
Key: 'path’ Value: './’
Binary Content *
Header
Content

User Interface
Less of this… … more of this

Deeper Ecosystem Integration: 274+ Processors,
57 Controller Services
Hash
Extract
Merge
Duplicate
Scan
GeoEnrich
Replace
ConvertSplit
Translate
Route Content
Route Context
Route Text
Control Rate
Distribute Load
Generate Table Fetch
Jolt Transform JSON
Prioritized Delivery
Encrypt
Tail
Evaluate
Execute
All Apache project logos are trademarks of the ASF and the respective projects.
Fetch
HTTP
Syslog
Email
HTML
Image
HL7
FTP
UDP
XML
SFTP
AMQP
WebSocket
Parse Records Convert Records

What Are the IoT Challenges?

IoT Challenges
• Limited computing capability
• Limited power/network
• Restricted software
library/platform availability
• No UI
• Physically inaccessible
• Not frequently updated
• Competing standards/protocols
• Scalability
• Privacy & Security
@_lennart

• When the Mirai attack has its own
Wikipedia page, that’s not good
• Hackers stole high-roller database from
casino via aquarium thermometer
connected to internet (04/2018)
Recent Examples

• Runs on JVM
• Provides UI for flow design & monitoring
• Security built-in
• TLS, authentication/authorization, encrypted data
• Handles practically any format/protocol
NiFi Solves Everything*

• NiFi supports AMQP, MQTT, UDP, TCP,
HTTP(S), CEF, JMS, (S)FTP, AWSIoT
• With a little pruning, NiFi can run on a
Raspberry Pi
NiFi for IoT

• NiFi is designed to “own the box”
• NiFi 0.7.x started up in about 10-15 minutes on RP3 (593 MB)
• NiFi 1.x started up in about 30 minutes on RP3 (760 MB)
• 33 new processors
• Rewrite for multi tenant authorization
• Complete UI overhaul
So Why Do We Need a Different Solution?

Enter Apache MiNiFi

• Get the key parts of NiFi close to where data begins and provide bidirectional
communication
• NiFi lives in the data center — give it an enterprise server or a cluster of them
• MiNiFi lives as close to where data is born and is a guest on that device or system
• IoT
• Connected car
• Legacy hardware
Apache NiFi Subproject: MiNiFi

• NiFi is big
• 1.7.1 release is 1.2 GB compressed
• Can be modified to run in restricted environments, but requires manual surgery
• Provides UI, provenance query, etc.
• Runs on dedicated machines/clusters — “owns the box”
• MiNiFi lives at the edge
• No UI
• 0.5.0 Java release is 67 MB, C++ release is 6.1 MB (0.2.0 fits on a floppy disk)
• “Good guest”
Why Build MiNiFi?

• MiNiFi Java (v0.5.0)
• Modified version of NiFi
• No UI
• YAML configuration
• Reduced processor count
• 63+ by default, more
available with
additional NARs
• MiNiFi C++ (v0.5.0)
• Written from scratch
• 33 processors by default
• Bi-directional site-to-site & provenance data
Flavors of MiNiFi

NiFi vs. MiNiFi Java Processes
NiFi Framework
Components
MiNiFi
NiFi Framework
User Interface
Components
NiFi

• NiFi
• Design flows
• Aggregate data from many
sources
• Perform routing/analysis/SEP
• MiNiFi
• Receive flows
• Collect data
• Send for processing
How Does MiNiFi Interact with NiFi?

• We’ve been imagining EDGE to CORE as a bi-directional linear system
• Let’s expand
that to the real
world
Let’s Add Dimensionality

• Data tagging/provenance
• Governance from edge (geopolitical
restrictions)
• Security (encryption, certificate-based
authentication)
• Low latency (immediate reactions &
decision-making)
What Does MiNiFi Provide? Connected Car Reference Platform Box
Tuner + DSRC CardConnectivity Card

MiNiFi on a Connected Car
Comprehension
Collection
CAN Bus
Gateway
MCU MCU MCU
Ethernet /
Ethernet AVB
Local
Interconnect
Network
Yet to be
established
protocol
Listen Ethernet Listen LINListen CAN Listen <>
Parse CAN Parse Ethernet Parse LIN Parse <>
Processing / Synthesis
Route
Transmit Execute PrioritizeFilter

MiNiFi on a Connected Car

• Site-to-Site
• NiFi protocol
• Two implementations
• Raw socket
• HTTP(S)
• Secured with mutual authentication TLS
• HTTP(S), (S)FTP, JMS, Syslog, File, Email, Process
MiNiFi Exfil

Edge Data Exploration

• IoT Device generating log messages
• Need to encrypt data on device
• Need to prioritize some data for
unreliable network connectivity
• Transmit data to central node
• Decrypt data and analyze
• Make determinations and modify live
flow
Scenario

• Simulate the log generation
• Schedule is customizable
• Script can write to dynamic location
NiFi as Test Harness/Environment

• Tails a log file
• Logs the raw contents (can be
multiple lines in time window)
• Splits into individual lines
• Filters the content
• Using parity of the timestamp
• Prioritizes
• Encrypts using AES/GCM
• Exfils to remote NiFi
Build the MiNiFi Flow

• Save as template from NiFi
• Run $ ./bin/config.sh transform
template.xml config.yml
• MiNiFi flow ready to run*
*Still need to set up TLS & encrypted properties
Export from NiFi to MiNiFi

• NiFi TLS Toolkit makes certificates &
keystores simple (and secure)
• Copy encrypted property value from
flow.xml.gz to config.yml (flow repo)
Setting up Crypto

• All data transmitted over TLS
is encrypted
• On NiFi, automatically
decrypted
• Attributes visible
• Content still encrypted
because of EncryptContent
processor
• Can serve as secure route for
follow-on systems
If We Really Have TLS, Why Encrypt?

• Receive the data over S2S
• Log the incoming messages
• Decrypt content
• Log again
Process Data In NiFi

Does It Work?

• Increase the write frequency
• Check that newer records (within tail
window) with higher priority arrive first
Prioritization?

• Window Aggregator
• If >60% odd in window, switch prioritization
• Encrypt with different keys for different tags & send to different follow-on systems
• Spotty network? Tell MiNiFi to cache low priority and send in batches
• MiNiFi rollover & pruning of monitored log
• Exfil MiNiFi provenance data to NiFi
Next Steps

Community

Community Example
• Jeremy Dyer
• Alexa + MiNiFi + Dyer 2.0
http://www.opensourcedad.com/apache/minifi-cpp/2016/12/18/poop-scale.html

What’s Next?

• NiFi 1.8.0 — … Oct 2018 (170+ Jiras)
• Jetty, DB improvements
• Auto load-balancing queues
• TLS Toolkit w/ external CA
• Record processor improvements
• MiNiFi C++ 0.5.0 — 6 June 2018
• MiNiFi Java 0.5.0 — 7 July 2018
• NiFi Registry 0.3.0 — 25 Sept 2018
Introducing Apache NiFi Registry
New Announcements

• Previously, flows were exported via XML
templates
• Didn’t contain sensitive values
• Couldn’t be updated in-place
• No tracking system
• NiFi Registry brings asset management
as first-class citizen to NiFi
• Flows can be versioned
• Flows can be promoted between
environments
Introducing Apache NiFi Registry 0.3.0
NiFi Registry for Dataflows

Community Health

Apache NiFi site
https://nifi.apache.org
Subproject MiNiFi site
https://nifi.apache.org/minifi/
Subscribe to and collaborate at
dev@nifi.apache.org
users@nifi.apache.org
Submit Ideas or Issues
https://issues.apache.org/jira/browse/NIFI
Follow us on Twitter
@apachenifi
Learn More and Join us

Thank you
alopresto@hortonworks.com | alopresto@apache.org | @yolopey
github.com/alopresto/slides

The First Mile - Edge and IoT Data Collection With Apache Nifi and MiniFi

More Related Content

What's hot (20)

Similar to The First Mile - Edge and IoT Data Collection With Apache Nifi and MiniFi (20)

More from DataWorks Summit (20)

Recently uploaded (20)

The First Mile - Edge and IoT Data Collection With Apache Nifi and MiniFi