Post on 29-Nov-2014
description
© 2014 Cisco and/or its affiliates. All rights reserved. 1 © 2014 Cisco and/or its affiliates. All rights reserved. 1
Michael Enescu CTO/Head of Open Source Initiatives
March 14, 2013
© 2014 Cisco and/or its affiliates. All rights reserved. 2 2
• How did we get here: virtualization
• What is Fog Computing and IoT
• Where are we going: the “Open” future of IoT
© 2014 Cisco and/or its affiliates. All rights reserved. 3
7.2 6.8 7.6
Rapid Adoption rate of digital infrastructure: 5X faster than electricity and telephony
50 Billion Internet connected things
50
2010 2015 2020
0
40
30
20
10
Bill
ions
of d
evic
es
25
12.5
Inflection point
Timeline
Source: Cisco IBSG, 2011
© 2014 Cisco and/or its affiliates. All rights reserved. 4 *Cisco VNI Study 2012
of “things” are unconnected
99%
Traffic Growth
4x Transition to Cloud*
Mobility
Wi-Fi 50% of Traffic (Video over Mobile Devices)*
The Network Has to
Change
Intelligent
Device Growth
2.5/Person BYOD
Programmable
Mobile and Cloud
Simple
The Network Is the Platform to Connect the Previously Unconnected
© 2014 Cisco and/or its affiliates. All rights reserved. 5
Internet of Things
Cloud
Any Device
Video
Virtual
Mobile Packet
Switched Routed
Network as Platform
Bridged
Unconnected DC
PC Voice & Data
Dedicated Fixed
Circuit Shared
The Network Is the Platform to Connect the Previously Unconnected
© 2014 Cisco and/or its affiliates. All rights reserved. 6
Centralized -> Decentralized
Decentralized -> Centralized
Fixed, role based model
Easier ops model, new apps
New devices, P2P, M2M
Dedicated compute loads
On-Demand, XaaS, AAA
New PIN’s, improved protocols
Mainframe & Client-Server
Virtualization & Cloud
IoT & Fog
1st 2nd 3rd Moore Nielsen Prediction
Pow
er +
com
pute
reac
h A
AA
Centralized -> Decentralized
© 2014 Cisco and/or its affiliates. All rights reserved. 7
• Storage and Compute declining faster • Network scales very differently than compute
Sensors will evolve faster than bandwidth Distributed computing more compelling over time
• Data gravity?
Computation
Storage
Communication
Moore’s and Nielsen’s predictions hold
© 2014 Cisco and/or its affiliates. All rights reserved. 8
1.1x109 dp/day Data points generated by sensors .5x1012 B/week
Data generated by an offshore oil rig
1x1012 B/day Data generated by an oil refinery
2x1013 B/hour Data generated by a jet engine
90% of the world’s data created in last 2 years IoT = Small sensors + Big Data + Action
IoT Traffic will grow at 82% CAGR through 2017* brings new dimensions we are barley beginning to sense
© 2014 Cisco and/or its affiliates. All rights reserved. 9
• Networking is changing: 50B+ Devices coming – Immense amount of data No longer about “data transport” Moving to “intelligence about data”: Understanding and taking actions
• Analytics are changing: Massive data => can not move data fast enough to analytics => move analytics to the data Real-time actions => processing compute closer to the source
• Consequently three new trends emerge: Dramatic growth in number of applications (+optimization/specialization) for analytics at the edge Dramatic growth in the computational complexity to ETL only essential data information to the core The drive to instrument the “data” to be “open” not closed/locked-in (more on this later)
Traditional model: Store First, Query Later
• Fetch, • Analyze, • Report
!
Generate Actionable Events, Integrate with Policy/Mgmt System
Store raw data or filtered data for further mining.
Data in Motion model: Process First, Store Optional
Input Data
Input Data
Rules can express: • Predicates and Filters • Contextual/Dimension Data • Aggregations • Pattern Matching • Categorization & Classification • Sub-queries …
• Fetch, • Analyze, • Report
Data-base waiting for Queries
Query-base waiting for Data
© 2014 Cisco and/or its affiliates. All rights reserved. 11
Time Present
Data Sources
M2M Gateway
Intermediate Server/Gateway
Data Center
Achieved Data
Achieved Data
Achieved Data
Data-Mining, Machine Learning, Pattern Recognition, Cause-Effect, etc.
Future
Predict
© 2014 Cisco and/or its affiliates. All rights reserved. 12
Cloud
Device/Smart Object
North/South Flows East/West Flows Fog
Fog Nodes can be multi-tenant Shared, public or private (like cloud)
Highly virtualized environment Secured & isolated tenants, QoS, workload distribution
Mixed ownership & operation Single entity, federation of agencies
Service Mobility Ability to migrate a running instance from cloud to edge
Fog is the distributed, hierarchically organized platform where the Internet meets the physical world at M2M scale
© 2014 Cisco and/or its affiliates. All rights reserved. 13
• Operating System: Linux – by far the most successful open source project ever
• Virtualization: Hypervisor: xen, kvm, … Network: OpenDaylight Controller, OVS, NFV, …
• Cloud: Open Stack, Cloud Stack, oVirt, …
• Applications: …
• Internet of Things: Eclipse M2M – over a dozen new projects: kura, krikkitt, … Linux Foundation AllSeen Multiple P2P
© 2014 Cisco and/or its affiliates. All rights reserved. 14
• “All IoT Software will be Open Source” • Why?
Open Source = collaboration (at the largest possible scale) Open Source = credibility Open Source dominates virtualization:
Hypervisor, OS, FS, … Cloud compute is a direct consequence of virtualization
Of the physical machine (compute, memory, storage, I/O), the physical network
Open Source dominates development environments, applications: Mobility, Data, Big Data, Analytics, Security, …
Fog as an extension of the cloud is no exception From gateways, to smart devices, do any device, to sensors Though more work and collaboration is needed
• And where is Open Source already in the Internet of Things?
© 2014 Cisco and/or its affiliates. All rights reserved. 15
@michaelenescu
Everywhere! Thank You