QoS-aware Network Operating System for Software Defined
Download
Report
Transcript QoS-aware Network Operating System for Software Defined
QoS-aware Network Operating System
for Software Defined Networking with
Generalized OpenFlows
Kwangtae Jeong, Jinwook Kim and
Young-Tak Kim
2012 IEEE/IFIP 4th Workshop on
Management of the Future Internet
1
Copyright © 2012, [email protected]
Outline
•
•
•
•
•
Introduction and related work
Architecture
The QoS-aware Network Operating System(QNOX)
Implementation and performance analysis
Conclusion
2
Copyright © 2012, [email protected]
Introduction and Related Work
• The SDN allows network operators to manage network
elements using software running on an external server.
• It splits the network into forwarding and control
elements, giving the operators more flexibility to
configure their network.
• The OpenFlow framework provides standardized open
protocol in switches and routers.
3
Copyright © 2012, [email protected]
Introduction and Related Work
• The network operating system (NOX) provides a
programming interface with high-level abstractions of
network resources.
• However, it fails in providing the necessary functions
for QoS-guaranteed SDN service.
• In this paper, they proposed a QoS-aware network
operating system (QNOX), providing QoS-aware virtual
network embedding, end-to-end network QoS
assessment, and collaborations among control
elements in other domain network.
4
Copyright © 2012, [email protected]
Architecture
• SE: Service Element, providing a user interface and
receiving the service request from users.
• CE: Control Element, end-to-end session control,
routing path establishment, flow table update.
• ME: Management Element, resource discovery, virtual
overlay network, performance monitoring.
• CKE: Cognitive Knowledge Element, mapping from
virtual network topology to substrate network topology.
• FE: Forwarding Element, e.g., a switch.
Copyright © 2012, [email protected]
5
Architecture
6
Copyright © 2012, [email protected]
QNOX – Resource Discovery
• When activating a new FE, an UPnP-based
protocol is used to automatically connect to ME.
• ME then informs CE, and then CE calculates a
new FIB (forwarding information base) and
updates and installs on every FE in this domain.
• Traditional shortest path spanning tree
algorithm.
7
Copyright © 2012, [email protected]
QNOX – Resource Discovery
8
Copyright © 2012, [email protected]
QNOX – Manipulating Service Requests
• The SE receives service requests with attributes of the
required QoS parameters(power of virtual node, delay,
jitter, packet error rate, packet loss rate…).
• SE checks and evaluates the availability of network
resources. If the requested QoS level is not available,
there may be some negotiation among SE and user.
• SE also contains the service life-cycle management for
the accepted services, and QoE/QoS monitoring
modules.
9
Copyright © 2012, [email protected]
QNOX – Inter-Domain Routing
• PCE, a Path Computation Element, RFC 4655.
• Backward recursive PCE-based path computation.
10
Copyright © 2012, [email protected]
Implementation and Performance Analysis
• Using modified Linux IP/MPLS router to emulate FE
and CE.
• Each element is individually running on a VMware
virtual machine. Each 19 FEs are grouped and
installed on a PC server. One of the 19 FEs in the
group is providing a direct link to the CE.
• One CE is configured for a domain network which
contains 2~114 FEs.
11
Copyright © 2012, [email protected]
Implementation and Performance Analysis
12
Copyright © 2012, [email protected]
Implementation and Performance Analysis
13
Copyright © 2012, [email protected]
Conclusion
• A clear architecture that can run on OpenFlow
and legacy network devices simultaneously.
• Need a mechanism that deals with the QoS
problem in runtime.
• Miss some details and issues in this paper.
• The evaluation is not convincible.
14
Copyright © 2012, [email protected]