0. Enhanced Network Control Plane architectures supporting Cloud Computing applications: GMPLS+/PCE+
Nicola Ciulli
Project Steering Committee
On-demand Infrastructure Services Provisioning Workshop,
OGF 28, Munich, Mar 15th2010

1. Contents A brief problem statement
NCP in the GEYSERS architectural layering
NCP main challenges
NCP major innovations
Any-IT + Network Service concept
Energy-efficiency
Achieving the GMPLS+/PCE+ NCP
Main work items
Starting points of work
Planned delivery of the major results
Expected impact on SDOs

2. NCP problem statement
Grids, cloud computing and SOAs at large rely on a vital commodity: the network
An ever-increasing number of distributed [super-]computing applications
Highly-demanding requirements for dynamicity and flexibility in Net + IT resource control (e.g. automated scaling up/down)
…but their network service(s) still treated as “always-on”
Application layer unable to exploit the automatic control potentialities of the current optical (and not-optical) network technologies
IT resources dynamics completely uncorrelated from the network ones
Common trend to over-provision network services
 inefficient resource utilization in the network, above all in case of fault recovery

3. NCP problem statement (cont’d)
[DEMAND] More and more massive utilisation of these Net+IT services by end-users
[OFFER] A rapid migration from telecommunication networks to global communication infrastructures
Network infrastructure
Computing/storage elements (IT)
Software and tools to control, monitor and manage the Net+IT services
Often virtually composed on top of incumbent providers
… and with a “revolution” in actors’ roles and responsibilities along the value-chain
Infrastructure providers, network operators, virtual network operators,
Service/Content providers, Over the Top (OTTs), etc.

4. GEYSERS reference model
Roles
Application
Application Interface
App/Service Providers
GMPLS+/PCE+
Network Control Plane (Extended ASON/GMPLS and PCE Control Plane)
(virtual)
Network
Operators
Logical Infrastructure Composition Layer
Explain all the concepts regarding diferent question:
Partitioning, multiplexing, SLA, energy efficiency, LICL, GMPLS +, PCE +,
a Holistic apporach that allow telco operators to go a to a new business model
Network operators will not only operate networks, but IT reosurces, so they need from the integrated framework
define storyboard (doc word)
Answer most of the quesitons inderectly by using this picture
define diferent types of users
The novelty in the GEYSER architecture lies in the following two layers: the Logical Infrastructure Composition Layer (LICL) and the Network Control Plane (NCP) layer. Under this architecture, several independent logical infrastructures can co-exist; isolated, yet sharing the same physical layer.
Logical Infrastructure Composition Layer: This layer allows/supports the partitioning of the physical infrastructure, including both Optical Network and IT resources. It utilizes a semantic resource description and information modelling mechanisms for hiding the technological details of the physical layer from network operators. Logical resources are represented seamlessly using a standard set of attributes which allows the Control Plane to overcome the network and technology segmentation. Partitioning provides a 1:N logical representation of a physical resource from one or multiple domains. The logical infrastructure layer allows for dynamic and consistent monitoring of the physical layer and binding/associating the right security and access control policies. Furthermore, this layer constitutes application specific logical infrastructures by interconnecting the logical resources based on applications’ requirements.
Network Control Plane: This layer is proposed as an extension of ASON/GMPLS and PCE, both in terms of architectural elements and protocol objects/procedures. The NCP layer is responsible for mapping applications requirements and control/management of the logical infrastructure composed by the LICL, seen and controlled just as a physical infrastructure. Each logical infrastructure can include both network and IT resources, and is controlled by a single instance of an NCP. The NCP is in charge of dynamic provisioning, monitoring and recovery functions. The main targeted features of the NCP layer are listed in the related objective in the previous section; these technical objectives will drive the specification of the NCP architecture and its prototype development.
As part of its architecture, GEYSER will define the workflow that provides full interoperability and interaction between infrastructure providers and network operators (with its application service requirements) for planning, provisioning and accessing network and IT resources.
The GEYSER architecture is also backward compatible and it is able to interact with its standard neighbouring domains at two levels:
Transit level: where the GEYSER Network Control Plane can interact with control and management of neighbouring domains to request or provide resources (e.g. transient bandwidth).
Hybrid infrastructure level, where the GEYSER Logical Infrastructure Composition Layer can interact with its neighbouring infrastructure domains to compose a logical infrastructure with the help of neighbouring infrastructures in offering their resources as a service.
Infrastructure Providers
Network Infrastructure
IT Infrastructure
Resources
IT: storage, computing
IT: storage, computing
optical infrastructure 4

5. NIPS: Network+IT Provisioning Service
Beyond a UNI, towards a App-to-Net i/f
Generalized semantics to describe
the functional characteristics of both IT sites and NEs (exported from LICL)
i.e. resource types, capabilities and availabilities
E.g. sites, attached services, capabilities and capacities of network, computing and storage elements, etc.
also the non-functional service characteristics
Workflow descriptions
Interaction properties
Service Level Agreements
Pricing/charging models
etc.
a service that allows the provisioning of network and IT resources in a single-step, through a set of seamlessly integrated procedures

6. Planned features [1]
Provide a single reference point for the Application layer (Network+IT Provisioning Service interface – NIPS)
[downward] e2e dynamic (advance) reservations with differentiated service guarantees (QoS but also resiliency) on a per-user/service basis
[upward] Reporting SLA fulfilment for subsequent workflow adjustments
Capability of dynamic e2e resource composition of Net + IT resources
“One-step” Net+IT resource reservation
Scheduled/advance reservation (various degrees from fully distributed to centralized)
Trusted mechanisms for accessing resources
Coordinated recovery strategies for an overall service resiliency

7. Planned features [2]
Supporting new connection paradigms, i.e. beyond p2p
assisted unicast (i.e. network quotations about performance and prices of <src, dst> pairs)
restricted anycast (i.e. with a set of equivalent dst in which NCP+ will pick the best one)
full anycast (i.e. dst choice totally delegated to NCP)
Coordinated service provisioning and maintenance through the LICL
Dynamic provisioning, monitoring and recovery functions
Logical infrastructure re-planning based on TE considerations, network & SLA fulfilment benchmarking
Backward compatibility and interoperability with std GMPLS and PCE
ASON/GMPLS (by ITU-T and IETF CCAMP WG)
PCE (by IETF PCE WG)
UNI and E-NNI interfaces (by OIF + possible influences from OGF NSI WG)

8. Reference deployment scenario
PCE+ protocol
PCE+ protocol
PCE
x-NNI
PCE+
NIPS UNI
PCE+
x-NNI
I-NNI
NIPS UNI
I-NNI
I-NNI
GMPLS+/PCE+
Region/Area
GMPLS+/PCE+
Region/Area
PCE+ protocol
Standard
GMPLS/PCE NCP
(inner core)
PCE+
x-NNI
I-NNI
GMPLS+/PCE+ NCP
(outer core)
GMPLS+/PCE+
Region/Area
NIPS UNI

9. NCP major innovations AAA for net+IT provisioning services GEYSERS
enhanced NCP
Application dynamics and granularity at the user-network interface
(NIPS)
Extended communication paradigms
assisted unicast
restricted anycast
full anycast
New service composition dynamics at the SOA layer
BoD services planned + provisioned by the NCP but optionally hocked to the NMS decision flow
Cross-layer service monitoring and escalation of recovery procedures
Scheduled BoD services handled by the NCP in a distributed/federated way
Energy-efficient network + IT resource routing and provisioning

10. Main work items for GMPLS+/PCE+
@Routing plane [GMPLS+/PCE+]
Protocol extensions for Net+IT resource advertisements
NIPS route computation
Procedures for inter-domain NIPS routing
@Signalling plane [GMPLS+]
Protocol extensions for Net+IT resource reservations
Procedures for seamless NIPS signalling
Procedures for coordinated inter-layer recovery and escalations
NIPS interface specification
To let end-user apps or the brokering/scheduling entities on the Enterprise Service Bus (ESB) stimulate the GMPLS+ NCP

11. Reference specs and architectures
ASON/GMPLS core specs
ITU-T SG15 work on ASON and Transport Ethernet
OIF for ASON Network Interfaces (UNI, E-NNI sig+rout.)
IEFT RFCs & I-Ds for
signalling (G.RSVP-TE)
routing (G.OSPF-TE)
recovery mechanisms
support for multiple switching technologies (MRN/MLN)
PCE architecture and inter-PCE comms
OASIS SOA application requirements and framework
OGF: NSI, OCCI?

12. Expected impacts on SDOs
e.g. by contributing to
NSI-WG
Tentative ISOD WG

13. Nicola Ciulli, n.ciulli<at>nextworks.it
Thank you
Any questions?
Nicola Ciulli, n.ciulli<at>nextworks.it 13

14. Energy-data awareness and crunching
Energy efficient approach
Logical Infrastructure Composition Layer
Partitioning and Virtualization
Statistically sharing of resources
Resource usage optimization
right-size infrastructure
Energy-aware
Scheduling
Allocation
Minimize the number of concurrent powered on resources
power down unused infrastructure
Extended Control Plane
Energy-aware Routing (GMPLS+/PCE+)
Power consumption ON + IT (metadata info) dissemination
create awareness
Energy efficient Routing and provisioning
prefer energy efficient resources
Promote Optical networking
Considerable energy saving compare to IP routers
traffic mapping on hybrid networks

15. Starting point of work: prototyping & testing
A number of existing platforms to be extended/used
For GMPLS+/PCE+ (protocol stacks in Linux controllers)
For std GMPLS/PCE (protocol stacks embedded in optical nodes)
… and a solid R&D test-beds background from the team
Telco SME: Nextworks (NCP WP Lead)
Telco manufacturers: ADVA, Alcatel-Lucent Italy
Pan-European Carriers: Interoute
Service Providers: SAP AG
Telecom Operators & NREN: Telefonica I+D, Polish PTT, PSNC
Universities and Research Centers: i2CAT, Univ. Essex, Univ. Amsterdam, Athens Information Technology, Tech. Univ. Braunschweig, IBBT 15 15

16. Planned delivery of the major results
[Q1-2011] GMPLS+/PCE+ architecture (high/low level)
[Q1-2012] First prototypes for testbed integration and validations
[Q3-2012] Final prototype releases