Date post: | 11-Apr-2017 |
Category: |
Education |
Upload: | hong-linh-truong |
View: | 342 times |
Download: | 0 times |
SINC – An Information-Centric Approach
for End-to-End IoT Cloud Resource
Provisioning
Hong-Linh Truong and Nanjangud Narenda
Distributed Systems Group, TU Wien
dsg.tuwien.ac.at/staff/truong
Ericsson Research, Bangalore, India
ICCCRI2016@CloudAsia2016, 4th May 2016 1
Outline
Motivation
Challenges
The SINC conceptual framework
Overall architecture
API management and integration
Naming, slicing, and routing
Slice management and adaptation
Towards the implementation of SINC
Conclusions and future work
ICCCRI2016@CloudAsia2016, 4th May 2016 2
State-of-the art IoT Clouds/Cyber-
Physical Systems Complex infrastructures of IoT elements (sensors,
gateways, networks), micro data centers, network
services, cloud VM, storage, etc.
ICCCRI2016@CloudAsia2016, 4th May 2016 3
Cloud
(big, centralized
data centers)
Cloud
(big, centralized
data centers)
Edge
(IoT devices, micro
data centers)
Edge
(IoT devices, micro
data centers)
Edge
(IoT devices, micro
data centers)
Edge
(IoT devices, micro
data centers)
Network functions
(network services
+ micro
datacenter
On-demand resources provisioning across IoT networks
(the edge), network functions (the middle) and the
clouds (the back-end)
Motivation (1)
Application scenarios: emergency responses, on-demand
crowd sensing, Geo Sports monitoring, cyber-physical
systems testing, etc.
ICCCRI2016@CloudAsia2016, 4th May 2016 4
Geo Sports: Picture courtesy
Future Position X, Sweden
Indian Overfly collapses
figure source: http://timesofindia.indiatimes.com
Need to have an end-to-end provisioning of resources
E.g., sensors, network function services, storage, virtual machines
Short, crucial and heavily workload; elasticity and uncertainties.
Motivation (2)
Problems
Virtual resources are provided by different providers
Often there is no coordination among them inadequate
support for elasticity and uncertainties for the application
Host-centric information is too low level to represent
“slice” view
It is very hard, if not impossible, to establish end-to end
view on resources
lack of tools, too complex, time-consuming, & error-
prone effort for application users and developers
Our contribution
A conceptual framework for slicing IoT, network
functions and cloud resourcesICCCRI2016@CloudAsia2016, 4th May 2016 5
Challenges
Modeling distributed IoT, network functions and cloud
capabilities in an integrated view
Slicing end-to-end network of resources
Composing resources in slices of IoT, network
functions and clouds
(Re-)configuring composed resources
ICCCRI2016@CloudAsia2016, 4th May 2016 6
End-to end
Resource slice
Applications/Virtual
infrastructures
SINC conceptual framework
ICCCRI2016@CloudAsia2016, 4th May 2016 7
Integrating diverse types of resources
Make a Resource Grid ready for slice creation
How to harmonize and gather IoT, network functions and cloud
resources
API Integration and Communication
Use REST API for obtaining metadata and control of resources
Sensoring data can be transferred through different
middleware
Work with existing metamodel (IoTivity, OpenHAB, IoTDM,
ETSI MANO, OCCI, CIMI, etc.)
Rely on scalable cloud middleware (e.g., AMQP & MQTT)ICCCRI2016@CloudAsia2016, 4th May 2016 8
IoT networks Network Function Services Clouds
Resource Grid
Naming, Slicing and Routing
From Resource Grid to information-centric description
of Partitions of Resources for slices
Information-centric description of resources from IoT, network
and clouds; modeling partitions of resources
Slicing
Leveraging network slicing techniques (e.g., 5G)
Leveraging IoT and cloud virtualization to provision on-demand
dedicated resources with elasticity capabilities
Routing
Utilize concepts of Forwarding Information Base (FIB) and
Pending Interest Table (PIT) for routing control commands and
data queries to underlying resources
Separate control commands and data queries from sensoring
data transportation
ICCCRI2016@CloudAsia2016, 4th May 2016 9
Resource Management,
Configuration and Adaptation (1)
Creating slices, each slice includes a set of partitions of
resources
Modeling and capturing user requirements for slices
Creation and Management
Develop new algorithms for creating slices by leveraging
existing works for IoT, networks, and services
Integrate with NFV orchestrators, virtual sensors, gateways,
cloud APIs and SDN controllers.
Deal with different resource provisioning models imposed by
underlying infrastructures
Configuration by leveraging different deployment tools for IoT,
network functions and clouds
ICCCRI2016@CloudAsia2016, 4th May 2016 10
Resource Management,
Configuration and Adaptation (2)
Monitoring and Management
Develop end to end metrics for slices
Integrate monitoring capabilities from different
providers and correlating monitoring data
Runtime slice adaptation
Performance as well as uncertainties at
infrastructures, applications and their integration
levels
Adaptation capabilities across IoT, network functions
and clouds
Multiple level of adaptations based on end-to-end
problems and partition problems
ICCCRI2016@CloudAsia2016, 4th May 2016 11
Towards the Implementation
Using REST API to integrate resource management
capabilities from different providers
Distributed communication middleware, e.g., based on
AMQP/MQTT, for querying resource information and
propaging controls
TOSCA or other topology description tools for modeling
topologies for supporting configuration and deployment
Leveraging existing deployment techniques for IoT and
clouds
Testbed established with open sources: Dockers,
OpenStack, Weave, OpenDayLight, etc. by utilizing
cloud, network and IoT devices
ICCCRI2016@CloudAsia2016, 4th May 2016 12
Towards the Implementation - HINC
Implement API Integration and
Communication
http://sincconcept.github.io/HINC/
High level information models for
Resource Grid
Middleware and adaptors for
integrating different providers
API for querying and configuring
resources
Leveraging SALSA for IoT,
network functions and cloud
configuration
http://tuwiendsg.github.io/SALSA/ICCCRI2016@CloudAsia2016, 4th May 2016 13
Conclusions and Outlook
Slicing IoT, network functions and clouds
Important for various types of applications
Key to the coordination of diverse types of resources in
distributed edge and cloud systems
SINC: a conceptual framework and steps to achieving end-to-
end resources provisioning
Ongoing work
Slice requirement modeling and composition algorithms
APIs for programming resource queries and controls
Configuration tools (http://tuwiendsg.github.io/SALSA/)
Uncertainty testing and analytics (www.u-test.eu)
Testbed (Vienna, Bangalore, Hanoi, and public clouds)
Check http://sincconcept.github.io for new update
ICCCRI2016@CloudAsia2016, 4th May 2016 14
Thanks for your
attention!
Questions?
Hong-Linh Truong
Distributed Systems GroupTU Wien
dsg.tuwien.ac.at/staff/truong
ICCCRI2016@CloudAsia2016, 4th May 2016 15