Software Defined Systems:

Software Defined Systems:
from SDN to SDI
ZKI Herbsttagung
24. September, 2014
Kaiserslautern
Paul Mueller
Integrated Communication Systems Lab
Dept. of Computer Science
University of Kaiserslautern
Paul Ehrlich Bld. 34, D-67663 Kaiserslautern, Germany
Tel.+49 631 205 2263, Fax. +49 631 205 3056
www.ICSY.de
Content
what we are talking about
the promise
ACI-REF
Paul Mueller, University of Kaiserslautern
2
we are talking about systems …
1
!  A system is a set of interacting components forming an
integrated whole of elements/components and
relationships.
!  Common characteristics are:
-  A system has a structure, it contains parts (or components) that
are related to each other;
-  A system has a behavior, it exhibits processes that fulfill its
function or purpose;
-  A system has interconnectivity: the parts and processes are
connected by relationships.
!  CS has to investigate the abstract properties of systems
matter and organization, looking for concepts and
principles that are independent of domain, substance,
type, or temporal scale.
1
Paul Mueller, University of Kaiserslautern
adopted from wikipedia
3
we are talking about systems …
1
!  A system is a set of interacting components forming an
integrated whole of elements/components and
relationships.
!  Common characteristics are:
-  A system has a structure, it contains parts (or components) that
are related to each other;
-  A system has a behavior, it exhibits processes that fulfill its
function or purpose;
-  A system has interconnectivity: the parts and processes are
connected by relationships.
Software)Defined.Systems.
provides.a.framework.to.
manage.complex.systems
!  CS has to investigate the abstract properties of systems
matter and organization, looking for concepts and
principles that are independent of domain, substance,
type, or temporal scale.
1
Paul Mueller, University of Kaiserslautern
adopted from wikipedia
4
demands
especially about IT-systems …
what is the right glue ?
eMail
Paul Mueller, University of Kaiserslautern
capabilities
WWW
5
what does it mean?
!  Application was integral
part of resource
-  Hard wired
!  Front end was separated
from backend Separation
of GUI and hardware /
logic (MVC)
!  Client / Server
-  N-tier modells
!  SDN/OpenFlow
-  Separation of data and
control plane
Paul Mueller, University of Kaiserslautern
6
what does it mean?
!  Application was integral
part of resource
-  Hard wired
!  Front end was separated
from backend Separation
of GUI and hardware /
logic (MVC)
Separation.of.concerns.
!  Client / Server
(Edsger W. Dijkstra – 1974).
-  N-tier modells
!  SDN/OpenFlow
-  Separation of data and
control plane
Paul Mueller, University of Kaiserslautern
7
more abstract ...
!  Applications
- 
- 
- 
- 
Facebook, Twitter, ...
SaaS, ..., IaaS
Industry
Cloud
Application / Services /
Processes
-  Hard wired
-  MVC
-  Client/server, …
!  Infrastructure
- 
- 
- 
- 
X
hard wired
!  Control
Control
SDMVC
client/server ...
Infrastructure
Compute
Storage
Network
Instruments
Paul Mueller, University of Kaiserslautern
8
more abstract ...
!  Applications
- 
- 
- 
- 
Facebook, Twitter, ...
SaaS, ..., IaaS
Industry
Cloud
- 
- 
- 
- 
Compute
Storage
Network
Instruments
Application / Services /
Processes
Software)Defined.Systems.
hard wired
!  Control
add.software.components.
Control SDMVC
-  Hard
wired
-  MVC
client/server ...
to.abstract.from.application.
-  Client/server, …
and.infrastructure.
!  Infrastructure
Infrastructure
Paul Mueller, University of Kaiserslautern
X
9
what have we learned from ...
!  GRID computing:
-  coordinates resources not subject
to centralized control using
standard, open, general-purpose
protocols and interfaces to deliver
nontrivial qualities of service.
-  GLOBUS toolkit, WSRF
-  FutureGrid (NSF project)
!  Cloud computing:
- 
- 
- 
- 
Virtualization
Hypervisor
OpenStack
Middleware (like Eucalyptus)
!  Service-oriented computing:
-  Resource description
-  Resource discovery
-  Resource binding
Paul Mueller, University of Kaiserslautern
10
1
Software-Defined Exchanges (SDXs)
"  “Near-term” SDX – pure connectvity/ROUTING
  Layer 3 (IP) – e.g., connect AS’s
  Layer 2 (Ethernet) – e.g., multi domain circuits
  SDN - connect SDN islands
"  “Advanced” SDX – with compute/storage (infrastructure)
 
 
 
 
Connect SDI island
Compute / storage / network / instruments (industrial robots)
Extend this idea to general systems
G-Lab / FIRE / GENI / ... as early instances
1
adopted from Workshop on Prototyping and Deploying
Experimental SDXs, Washington DC, June 2014
Paul Mueller, University of Kaiserslautern
11
What does SDX exchange?
!  Application data? IP
Packets? Virtual
topologies? ….
!  While SDX’s can be
defined broadly
(everything under the sun!)
!  At the base level, SDX is a
more intelligent IP
exchange
!  At the sophisticated level,
SDX is a virtualized,
application-specific
exchange
!  Building the right
framework for an
exchange will be critical to
support all definitions
1
adopted from Workshop on Prototyping and Deploying
Experimental SDXs, Washington DC, June 2014
Paul Mueller, University of Kaiserslautern
12
SDX ?
!  A place where peering domains
come together based on SDX-defined rules of engagement
!  Supports establishment of inter-domain connectivity/routing and/or
SDSs by applications – supports
exchange of information needed to
achieve these
!  Enforcer of individual domain policies
but not itself a policy definer (with
exceptions)
!  A broker of inter-- domain and
inter-- SRP trust
!  A marketplace for services – local
and remote
!  Peering domains need not be
physically connected
!  Variation/differentiation among
SDX’s is useful, but some common
services are necessary
!  Increasing complexity of global
multi-- domain computing /
communications environments
!  New paradigms such as SDN, SDI,
(distributed) clouds, virtualization/
slicing, big data
!  Applications that need to utilize
resources in multiple domains
-  Science requirements from different
discipline groups – workflow,
requirements, and network needs for
these
-  For R&E, end to end is important –
e.g., “hand fashioned circuits for
astronomy”
!  Declarative control of inter-domain
path end to end – BGP doesn’t do
this
!  Potential to leverage awareness of
multi-domain network conditions
!  Security included in the infrastructure
!  Rich environment for researchers to
control their experimental
infrastructure needs
1
adopted from Workshop on Prototyping and Deploying
Experimental SDXs, Washington DC, June 2014
Paul Mueller, University of Kaiserslautern
13
SDN stiching
1
adopted from Workshop on Prototyping and Deploying
Experimental SDXs, Washington DC, June 2014
Paul Mueller, University of Kaiserslautern
14
SDN stiching
Paul Mueller, University of Kaiserslautern
15
The Project: ACI-REF
!  Based on OpenGeni racks
- 
- 
- 
- 
DELL
Omnibond (spinoff University Clemson)
GENI project office (GPO)
NSF project ACI-REF
•  Advanced Cyberinfrastructure –
Research and Educational
Facilitation: Campus-Based
Computational Research Support
!  Service registration and discovery
(what we have learned from SOA)
!  Plug-in services
- 
- 
- 
- 
Data transfer node, high throughput
storage
Application specific services (e.g.
agents installed and launched on-thefly applications)
Authentication software
Relocation tranparency of VMs
!  Interfaces for ...
- 
Applications to inquire and specify
resources
!  Marketplace of SDX services
- 
Self-service, on-demand services
instantiated by applications
!  Deployment
-  Software installation on GENI
racks
-  Network configuration across
campuses, regional, national,
international
!  Bring up specific pilot
applications
-  Can be partly manual
-  Identify resource and attributes
-  Identify policies (with network
operators)
-  Multi-domain provisioning and
policy enforcement
!  Integrate core services
-  Service registration & discovery
-  AAA
!  Self service interface to push
button deploy services into SDX
!  Analyze trust/security issues at
various levels of SDX hierarchy
•  Push button deployment
•  Application registration & discovery
(incl. authentication and ...)
Paul Mueller, University of Kaiserslautern
16
ACI-REF: KL – Clemson SDX
Paul Mueller, University of Kaiserslautern
17
ACI-REF: KL – Clemson SDX
Paul Mueller, University of Kaiserslautern
18
KL – Clemson SDX: DELL OpenGeni Rack
!  OpenGENI™ Network Connectivity
- 
- 
- 
- 
- 
- 
The OpenGENI rack implements
physically separate network planes to
improve system reliability, enhancereproducibility, and segregate
experimental and management traffic.
Control Plane: OpenFlow and GRAM
commands are passed between the
Controller and Compute nodes on this
network. This is not controlled by
OpenFlow and has its own internal IP
addresses for each node.
Data Plane: The OpenFlow-controlled
network and associated interfaces
between the VM’s representing the
requested compute/network resource
topology.
External Plane: This network connects
the Head node and switches to the
external internet.
Management Network: This network
enables SSH entry and between the
created VM’s. This is not controlled by
OpenFlow.
iDRAC: Integrated Dell Remote Access
Control. iDRAC helps to deploy,
update, monitor and maintain Dell
servers.
Paul Mueller, University of Kaiserslautern
19
ACI-REF: KL – Clemson SDX
!  What have we learned from distributed systems?
-  Migration transparency
-  Relocation transparency
•  Relocation of VMs
SDX
Kaiserslautern
X
Paul Mueller, University of Kaiserslautern
Clemson
… other
CPU, RAM, I/O
20
ACI-REF: KL – Clemson SDX
-  Where is my virtual machine?
-  Can I move my environment variables (like IP, MAC,
QoS, ...) also?
-  are a set of dynamic named values that can affect the
way running processes will behave on a computer.
SDX
Kaiserslautern
X
Paul Mueller, University of Kaiserslautern
Clemson
… other
CPU, RAM, I/O
21
Conclusion
!  What is missing for realizing the vision of SDS/SDI?
!  Need of a new layer
-  of abstraction above hardware resources (southbound API)
•  e.g. OpenFlow
-  of abstraction below application layer (northbound API)
!  Layer allows to empower application to define their IT
demands / environment
-  so that sufficient capacity to fulfill QoS/QoE appropriately
-  based on performance, security and availability requirements
!  Possible solution for this layer
-  use SOA paradigm
-  all infrastructure is virtualized and is delivered as a service.
-  connecting applications with infrastructure elements is completely
automated by software
!  Need for adjusting paradigm
-  new languages for describing service and service contracts
-  new technologies for accessing services
!  ACI-REF across Germany, Europe, …
Paul Mueller, University of Kaiserslautern
22
demands
what we are talking about …
eMail
Paul Mueller, University of Kaiserslautern
capabilities
WWW
23
Prof. Dr. Paul Mueller
Integrated Communication Systems ICSY
University of Kaiserslautern
Department of Computer Science
P.O. Box 3049
D-67653 Kaiserslautern
Paul Mueller, University of Kaiserslautern
Phone:
Fax:
+49 (0)631 205-2263
+49 (0)631 205-30 56
Email:
Internet:
[email protected]
http://www.icsy.de
Literature
!  [1] Paul Müller, Bernd Reuther: Future Internet Architecture - A Service Oriented Approach. In: it - Information Technology, Jahrgang 50 (2008) Heft 6, S.
383-389 6/2008 Link: http://www.oldenbourg-link.com/doi/abs/10.1524/itit.2008.0510
!  [2] Dennis Schwerdel, Zornitsa Dimitrova, Abbas Siddiqui, Bernd Reuther, Paul Müller: Composition of Self Descriptive Protocols for Future Network
Architectures. In: 35th EUROMICRO Conference on Software Engineering and Advanced Applications, Patras, Greece, September 2009, p. 585-591, Link:
http://dspace.icsy.de/handle/123456789/255
!  [3] Dennis Schwerdel, Daniel Günther, Robert Henjes, Bernd Reuther, Paul Müller: German-Lab Experimental Facility. Future Internet - FIS 2010, Lecture
Notes in Computer Science, 6369, 2010, Link: http://dspace.icsy.de:12000/dspace/handle/123456789/306
!  [4] M. Rahamatullah Khondoker, Bernd Reuther, Dennis Schwerdel, Abbas Siddiqui, Paul Müller: Describing and Selecting Communication Services in a
Service Oriented Network Architecture. In: In Proceedings of the ITU-T Kaleidoscope event, Pune, India, 13-15 Dec 2010, Link:
http://dspace.icsy.de/handle/123456789/287
!  [5] M. Rahamatullah Khondoker, Eric Veith, Paul Müller: A Description Language for Communication Services of Future Network Architectures. In: Proceedings
of „Network of Future“ (NoF), Paris, France, 2011 (nominated for best paper award), Link: http://dspace.icsy.de:12000/dspace/handle/123456789/325
!  [6] Khondoker, M. Rahamatullah, S. M. Taslim Arif, Nathan Kerr, and Dennis Schwerdel. Self-organizing communication services in future network
architectures. In Fifth Internationl Workshop on Self-Organizing Systems (IWSOS 2011), Karlsruhe, Germany, February 2011, Link:
http://dspace.icsy.de:12000/dspace/handle/123456789/296
!  [7] Dennis Schwerdel, Daniel Günther, Khondoker, M. Rahamatullah, Bernd Reuther, and Paul Müller. A building block interaction model for flexible future
internet architectures. In 7th EURO-NF CONFERENCE ON NEXT GENERATION INTERNET, June 2011, Link:
http://dspace.icsy.de:12000/dspace/handle/123456789/304
!  [8] Dennis Schwerdel, David Hock, Daniel Günther, Bernd Reuther, Phuoc Tran-Gia, Paul Müller: ToMaTo - a network experimentation tool. 7th International
Conference on Testbeds and Research Infrastructures for the Development of Networks and Communities TridentCom 2011, Link:
http://dspace.icsy.de:12000/dspace/handle/123456789/309
!  [9] M. Rahamatullah Khondoker, Abbas Siddiqui, Bernd Reuther, and Paul Müller. Service orientation paradigm in future network architectures. In Sixth
International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (IMIS-2012), July 2012, Link:
!  [10] Daniel Günther, Dennis Schwerdel, Abbas Siddiqui, M. Rahamatullah Khondoker, Bernd Reuther, Paul Müller: Selecting and Composing Requirement
Aware Protocol Graphs with SONATE, 12th Würzburg Workshop on IP: ITG Workshop "Visions of Future Generation Networks" (EuroView2012), Link:
http://dspace.icsy.de:12000/dspace/handle/123456789/362
!  [11] Dennis Schwerdel, Bernd Reuther, Thomas Zinner, Paul Müller and Phuoc Tran-Gia. Future Internet research and experimentation: The G-Lab approach.
In Computer Networks, Available online 8 January 2014, ISSN 1389-1286
!  [12] Paul Müller, Dennis Schwerdel and Justin Cappos. ToMaTo a Virtual Research Environment for Large Scale Distributed Systems Research. PIK - Praxis
der Informationsverarbeitung und Kommunikation, Available online 25 January 2014
Paul Mueller, University of Kaiserslautern
25