TERENA Networking Conference 2008 Workshop on «Challenges in Network Control and Service Planes»

Transcription

1 CAlcul Réparti sur Réseau Internet Optique à CApacité Surmultipliée* *Grid Computing services offered by Internet Optical Network with Ultra-high transmission Capacities TERENA Networking Conference 2008 Workshop on «Challenges in Network Control and Service Planes» «CARRIOCAS*: Orchestrating dynamic network service deliveries over ultrahigh capacity optical networks» Olivier Audouin, Olivier Leclerc, Dominique Verchere, Lionel Tuhal, Dominique Rodrigues Alcatel-Lucent Bell Labs France May 19 th 2008 BELNET - Brugge Belgium Vx/N&N/E/07/xxx 1

2 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 2

3 CARRIOCAS Challenges To research, design and implement a ultra-high bit rate network reconfigurable according to grid applications connectivity requirements Transmission capacity 40Gbit/s per wavelength Enhanced application-network interactions with SRV Guarantied network service deliveries in function of Application QoS requirements. To develop new Services & Usages collaborative engineering on virtual prototypes Computing Intensive Simulation Applications interactive visualization To specify new business models of services integrating IT access and connectivity 3

4 CARRIOCAS Partnership Optics Networking Computing Application Industrials SME s Academia Oct 2006-Sept 2009 LIST Financial partners: 4

5 CARRIOCAS Projects Organization Project Interactions: year1 year2 year3 Architectures Specification Reconfigurable Optical Nodes (QoS aspects at 40 Gbps) Network Requirement Analyzes of Grid Applications Project 1 Ultra-high Capacity Transmission Systems Engineering Project 2 Network Infrastructure Management and Control R. & D. Project 4 Distributed Computing & Storage Application R. & D. Applications/Networks Interactions Specification 40Gbit/s Transponder 40 Gbps Systems Engineering Experimentation Plan & Tests on Pilot Network Scheduling Reconfiguration & Virtualisation Prototype (interfaces) Experience and Test results on Pilot Network First tests of Distributed Applications Grid Computing, Storage & Visualization tests and Validation SP3 Pilot Network Operation 40 Gbps deployment and demonstration 5

6 A Collection of independent Resource centers: Network + IT infrastractures Collaboration between PoP is restricted to none stringent grid applications Total IT capacity can be very large but globally inefficient because collaboration restrictions due to network resource capabilities (limited performances) Some Datacenter systems can be overloaded while others are idle Company B Company A High-Capacity transmission: 10 Gbps per λ Datacenter B Capacity B Datacenter A Capacity A High-Capacity transmission: 10 Gbps per λ High-Capacity transmission: 10 Gbps per λ High-Capacity transmission: 10 Gbps per λ Company C Company E High-Capacity transmission: 10 Gbps per λ Datacenter E Capacity E Company D 10 Gbps per λ Datacenter D Capacity D Datacenter C Capacity C 6

7 Access of remote IT resource systems (Computational/Storage) at Datacenter Collaboration between Companies improved thanks to Network infrastructures IT resource usage efficiency is improved since jobs are processed faster End-to-end network + IT processing latency is minimal Network + IT resource sharing is maximized Company A High-Capacity transmission: 40 Gbps per λ High-Capacity transmission: 40 Gbps per λ High-Capacity transmission: 40 Gbps per λ Company B High-Capacity transmission: 40 Gbps per λ Company E High-Capacity transmission: 10 Gbps per λ 10 Gbps per λ Company C Company D Datacenter Capacity (A+ B+C+ D+ E) 7

8 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 8

9 Needs for Interactions for suitable Network Service provisioning Grid applications are overlaid and static connection: Dedicated infrastructures or over-provisioning are required to sustain the highest value of the connectivity requirements needed Shared resources and no control on the requirements (best-effort) No interactions Network (re)configurations are achieved manually Enhancements of Grid Application Network interactions to: Allow network management/control plane to deliver the right connectivity service with the correct requirements (performance, availability, security) Configure Network according to application workflows, (re)configuration should be fast, accurate and automatic (can not be manual). End-to-end Orchestration for better resources usages where crossoptimization can be deployed 9

10 Distributed Application Classification according to Network Requirements Distributed Applications Application with strict Reservation Data Transfer Application type Collaborative Application Information based Application Messages based Application Hard Real Time Resources always Required Soft Real Time Resources always Required Hard Real Time Soft Real Time Resources Sporadically Used Resources Sporadically Used Resources Used when Available Continuous Transmissions strict time constraints Continuous Transmissions loose time constraints Intermittent Transmissions strict time constraints Intermittent Transmissions loose time constraints Intermittent Transmissions when available Permanent & Strict Guaranteed Network Services SP4 (Visu( Visu.. Haute Perf.) Pharmacy (MED-Sumo) Energy (EDF) Telemedicine/Surveillance Critical supervision Commercial Grids Financial Data Storage Medical Data Storage SP4 (LUSTRE) Info Ctrl RPC Data Bulk Transfer Data Grid (LHC, LOFAR) Emergency Communication Gaming/Gambling Statistical Guaranteed Network Services Web-Browsing, Video-conf Peer-to-Peer file exchanges «Best-Effort» Network Services s, ftp Telnet sessions Asynchronous transfer (YouTune) 10

11 Distributed storage of massive volumes of data LUSTRE distributed file systems over CARRIOCAS metropolitan-scale network Cliquez pour modifier Security & le style du titre Access Control Clients Resource Database Metadata Control (MDS) Coherence Management Data Transfers Storage Management 2 types of flows Object Storage Targets (OSS) Control messages: Lot of requests, Low volume of data, end-to-end latency constraints Bulk payload data transfers: low number of requests, large volume of data 11

12 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 12

13 CARRIOCAS Network services suited for Grid Applications Connectivity services Requirements: to guarantee Quality of Experiences for grid application users Adaptive connections configurations with QoS guarantee (bandwidth, latency, availability ) negotiated for Grid Application workflow needs Technology enabler: GMPLS, dynamic SLA management Full virtualized services Requirements: To manage grid resources for executing the Grid Application workflows Selection, reservation and interconnection of Computational, Storage and Visualization Centers which may come from 3 rd party resource infrastructure providers (e.g. datacenters, HPC-centers) Co-selection and cross-optimization of network and grid (application) resources Technology enablers: conjunction of OGSA and network protocols Virtual Reality Picture Walls Computer Cluster Shared Grid Applications Optica l Tra nsport N etwork Grid Application Resources (server side) Da ta Center Instrument(s). Storage Center Local Networks. Grid Application (user side) Access Core Nodes Edge Nodes Computing Server HPC Computing 13

14 Extreme Data Flows require Carrier Grade Networks services Visualization pipeline partitioning and mapping on Grid IT Resources Data Source Cluster File System Raw data buffer Volumetric transfer function, Cliquez pour Surface modifier construction le style du titre 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D 3D Transformed Filtered data Data transformation: data Frame Filtering Rendering Display 20 Mpixels x 32 bits x 30 images/s x 2 for stereo = 38,4 Gb/s Low latency requirements for interactivity One Data flows can be mapped on one EVC allocated on dedicated WDM Auto-provisioned end-to-end Ethernet Virtual Circuit Dynamic mapping (CE-VLAN) TSS Add/Drop Ports PSC TE Link WDM FA-LSP WDM LSP Mapping (CE-VLAN) TSS Add/Drop Ports PSC OCh OCh PSC [PSC,WDM] TE-link Transport Service Switch (Ingress ROADM) [WDM,WDM] TE-link [WDM,WDM] TE-link Automatic Switched Optical Network (Egress ROADM) [WDM, PSC] TE-link Transport Service Switch 14

15 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 15

16 Ingress edge node LSR1 Packet LSP Routing Domain Carrier Grade Ethernet Network Working Packet LSP Transport Network Planning: TSS and OXC port dimensioning Cliquez pour modifier LSR2 le style du titre Backup Packet LSP Virtual Topology CORE Egress edge node Transport Service Switch GMPLS X-Connect Ethernet / T- MPLS λ / GMPLS Ingress core node Working Lambda LSP Wavelength X-Connect CORE Intermediate edge node Backup Lambda-LSP UNI Egress core node Routing Domain Oslo Stockholm Glasgow Copenhagen Dublin 3 41 Hamburg 38 Amsterdam 30 London Berlin Warsaw Brussels 27Frankfurt Prague Paris Strasbourg 12 Munich Zurich 13 Budapest Vienna Bordeaux Lyon Zagreb Belgrade Milan Madrid 9 Rome Barcelona PoPs (TSS + X-Connects), 10 Gbps / 40 Gbps interfaces 43 WDM Links, Network Connectivity d 3.1 COST266 network reference Traffic Matrix: from 5.26 Tbits on 2005 to 34.5 Tbits on 2010 with different Traffic Flow entries: Voice data, enterprise transactional data, datacenter grid data, internet residential data Athens 16

17 Point-to-Point connections (P2P): Ethernet Virtual Private Line services P2P LSP is composed by different nodes: 1 Ingress, 1 Egress, 0, 1 or many Transit nodes. Traffic is never replicated. Constraints are based on P2P LSP such as bandwidth requirements, resource class affinities, fast-rerouting and pre-emption. End-to-end delay requirements have to be satisfied for each LSP from the source to the receiver. Source Ingress1 Receiver Edge Nodes Receiver Egress1 B1 B2 Egress4 Receiver Egress2 X4 Core Nodes X3 Egress3 Receiver Receiver B1 and B2 are branch nodes 17

18 Point-to-Multipoint connections (P2MP): Ethernet Virtual LAN Services P2MP LSP is composed by different nodes: 1 Ingress, N Egresses, Branch nodes, Transit nodes. Branch nodes replicates traffic. X4 is a bud node i.e. it is Egress node and has one or more directly connected downstream LSRs Constraints are based on P2P LSP such as bandwidth requirements, resource class, fast-rerouting and pre-emption. Path constraints have to decide which LSR should be used as branch LSRs considering the node capabilities to replicate traffic. Source Ingress1 Receiver Edge Nodes Receiver Egress1 B1 B2 Egress4 Receiver Egress2 X4 Core Nodes X3 Egress3 Receiver Receiver B1 and B2 are branch nodes, data replication occurs 18

19 Solution Scalability of a solution from 2005 to 2010 Costs Growth % 100% 95% 90% 85% 80% 75% 70% 65% 60% 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% IP/Pos over PXC over Transparent WDM with 1850 TSS - LL=STM-64 LH - IL=10GE 200m reach Year Network level study Synthesis of some of the simulation results Comparison of all the solutions per year Layers Costs per solution Carrying cost per Gbps for all solutions from 2005 to IPPXCW_L10TS IPMETHW_L10 IPSXCW_L10TS ETHW_U10 IPPXCW_L10RT IPSXCW_L10RT Gbps carrying Cost of all solutions - Comparison over several years IPW_U10TS Layers Costs IPW_U10RT LL cost growth % IL cost growth % Total Traffic Growth % All Layers cost growth % Low Layers Cost Intermediate Layers Cost Upper Layers Cost All Layers Cost Low Layers Cost Intermediate Layers Cost Upper Layers Cost All Layers Cost Year ETHW_U10 IPMETHW_L10 IPPXCW_L10RT IPPXCW_L10TS IPSXCW_L10RT IPSXCW_L10TS IPW_U10RT IPW_U10TS 19

20 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 20

21 Different Options for Grid Application / Network interactions Compendium of architectures with three alternatives: Case 1: Application - Network Resource Management are disjoint: ACHIEVED End-Hosts solely drive the Grid applications operations on full provisioned and static networks performance rarely match Application requirements Restricted level of automations of the Network configuration DEVELOPMENT Case 2: Management Gateway : GridMS NMS interactions Abstracted models: Public + Restricted info/data models Connection requests are sent to centralized Service component Network service optimization at different granularities (e.g. L1 and L2) Rules / Policies defined based on Service Requests and Deliveries RESEARCH Case 3: Application Network Control functions interactions: Workflows / Connection Controls are processed by Active nodes Full exchanges of Control/Management info enabling X-optimization Require new contributions to standardize extensions of protocol engines 21

22 Case 2: Applications Interceptions for suitable Network Service configurations Enhancements of Grid Application Network interactions : management/control plane to deliver the right connectivity service Configure Network services according to Grid application needs, (re)configuration should be fast, accurate and automatic Orchestration for resources usages where optimization is envisioned Grid Application Users Request Storage Cluster of Enterprise B network interceptions Source UNI Path(ERO,label,) (3) TE Link Resv (8) Response Intermediate SOAP/HTTP/TCP/IP Path(ERO, label, ) (6) Resv (7) Destination interactions Path(ERO, label,...) (1) Path(ERO, label, (4) Resv (2) Resv (5) 22

23 CARRIOCAS Network Control & Management Architecture GUI Grid Application Application Requirements Scheduled Network Configuration Resource Co-Reservations Grid Resource Repository Agreement Data Center: computing Company A Abstract view Cliquez pour Application modifier Interface le style du titre Application Control interfaces GRAAP, DAI, BPEL/SOAP Grid Application resource Manager / Controller Discovery Grid Resource Network Interface System Architecture & Experiments Connectivity Requests: SLS, Negotiation, Policies & Rules Transport Ethernet T-MPLS (L2-VPN) Application Middleware Signaling UN I-N Grid User Network Interface (XML based protocols ) Scheduling, Reconfiguration & Virtualization (SRV) SLA publication SLS template Reachability / Virtual Topology Web Service Interfaces UN I-N (G)M PLS controlled N etwork Mediator Edge Node Network resource Manager / Controller Discovery Policies / Rules Software Architecture & Experiments Resource Center: Visualization Edge Nodes Signaling of application workflow Design/Development Deployment Data Center (DC) Abstract view of network topology and status Network Resource Repository 40 Gbps System Design D2.9 Grid Resource Network Interface 23

24 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 24

25 from «Internet Service Providers» to «Grid Service Providers» Entities for Dynamic SLA Negotiation and Service establishments Network Operations, Grid Resource Management, Grid Service Providers and Grid Application Users Interactions with Network Operators through a defined Service Interface. High Energy Physics Bio-informatics Sciences Applications Grid Application Users (Universities / Research Labs) Usage SLA Service Integrator & Presentation Service Storefront Grid Service Providers SLA Provider (Network Services) Specific Software (Resources) Application Service Computing Service Storage Service Visualization Service Acquisition Service Network Service CPU Resources Storage Resources Computational Fluid Dynamic HPC center Service Management Systems Interface (SML-i/f) Network Management / Control functions UN I-N UN I-N Provider SLA Picture Walls (Resources) Scientific Instruments/Sensors (Resources) Data Center (DC) Customers (Service Consumers) (G)M PLS controlled N etwork Mediator Edge Node Network Operator (Network Resource Supplier) Edge Nodes Acquisition Center 25

26 Specifications for Applications to interact with Grid Networks (Case 2) Scheduling Reconfiguration and Virtualization (SRV) Shared carrier networks can get or infer the connectivity service requirements from Grid Service Provider (workflow based) Management Gateway for Service Discovery, Selection and Signaling Just-In Time connection service provisioning (with correct QoS parameters) Arbitration of Network Resource Reservations for Services deliveries Virtualization of Network Resources to reduce management complexity Level of abstraction can be dependent of Grid Service Providers and Users End-to-end resource cross optimization can be enabled Grid Management System Network Management System 26

27 SRV Position and Interfaces (phase 1) Scenario1 : Network Services requests Grid Application User SLA Usage (Grid Services) Interfaces with Network Management System, SRV has access to all network information, SRV is an extension of the NMS SRV provides the Network Services according to the Architecture of the Grid Services: OGSA Grid Service Provider (Meta-Scheduler) constraints Grid Application User Grid Application User Grid Application User (e.g. Company A) SLA Usage (Grid Services) constraints Grid Service Provider (Meta-Scheduler) SLA Usage (Grid Services) constraints Grid Service Provider (Meta-Scheduler) SLA Usage (Grid Services) constraints Grid Service Provider (Meta-Scheduler) SLA Provider (Network Services) Negotiation/Agreement/Notification SLA Provider (Network Services) SLA Provider (Network Services) Negotiation/Agreement/Notification SLA Provider (Network Services) Service Layer SRV Connectivity Services SRV Connectivity Services Network Layer Network Management System TE-DB Controller PCE Controller Data Center Network Management System TE-DB PCE Controller Controller Controller Data Center Controller Controller Network Domain 1 Controller Data Center (e.g. Computing + Storage) Controller Controller Network Domain 2 Data Center Data Center 27

28 Scenario2 : Virtualized Service requests SRV Interfaces extensions towards the Datacenter (phase 2) Interface with Grid + Network Management System (GMS + NMS), SRV has access to Resource information of GMS and NMS through an extended interface SRV positions the Virtualized Services within the Architecture of the Grid Services: OGSA Grid Application User SLA Usage (Grid Services) constraints Grid Application User Grid Application User Grid Application User (e.g. Company A) SLA Usage (Grid Services) SLA Usage constraints (Grid Services) constraints SLA Usage (Grid Services) constraints Grid + Network Layer Service Layer Business Layer Grid Service Provider (Meta-Scheduler) SLA Provider (Virtualized Services) Grid Management System GE-DB Data Center Grid Services Grid Resources request/accept Controller MPLS Negotiation/Agreement/Notification EMS SRV Network Domain 1 Grid Service Provider (Meta-Scheduler) Network Resources request/accept Network Management System engine engine EMS SLA Provider (Virtualized Services) Connectivity Services engine PCE TE-DB engine engine Data Center Data Center (e.g. Computing + Storage) Grid Service Provider (Meta-Scheduler) SLA Provider (Virtualized Services) Controller GMPLS Negotiation/Agreement/Notification engine engine SRV Connectivity Services Network Resources request/accept Network Management System engine engine PCE Network Domain 2 engine Grid Service Provider (Meta-Scheduler) SLA Provider (Virtualized Services) TE-DB Grid Management System GE-DB Data Center Data Center Grid Services Grid Resources request/accept 28

29 The three main layers SRV components and its interfaces Grid Service Provider (Meta-Scheduler) Grid Application User Ext. OSS/BSS Charging/Billing Service Service Service Negotiation Service Publication Publication Negotiation S.R.V. // Notification Notification Patterns Policies Patterns Policies (WSLA) Contracts (WSLA) Contracts Service Service Selection Selection Scheduling Composition Scheduling Processes Processes Composition Shared Shared Information Information // Data Data Model Model Resource Resource abstraction abstraction // Virtualization Virtualization Web Service interface / OGSA compliant Mediation Network Management interface Management Engine: NMS Data/Info model Mgt Info Base (MIB) SNMP / Connection mgt Network Layer Control Engine: (G)MPLS Signaling: RSVP-TE Routing: OSPF-TE TE-DB Policy Path Computer Engine: PCE Policies / Rules PCE protocol 29

30 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 30

31 References on Experimentation plans Ethernet MPLS-TP over WDM Network architecture referencing : Carrier Grade Ethernet based Transport Networks: 1850TSS Network Configurations based on Management Functions: 1354BM-ETH service functions Large Scale Distributed Applications (Grid Applications): Computing and Storage Service on Demand (Ter@tec): LUSTRE CFS Application Instruments: Industrial (Remote Visualization): EDF Picture Walls Scheduling Reconfiguration & Virtualization (SRV) (SRV) service Negotiation protocol (e.g. WS-agreement based) Network Services Advertisement protocol (e.g. WS-notification based) North Bound Interface South Bound Interface Network Management Systems (1353NM 1354BM-Ethernet) Application Client / Server Network Controller Grid Application Client / Server Ethernet Switch (Customer Equipment) EDF Clamart (PoP 1) 1850 TSS-320 (Provider Edge) 1626LM Network Controller High-Capacity transmission λ: nx10 Gb/s link 1 Network Controller CEA-Saclay (PoP 3) 1626LM PE Ethernet Switch (Customer Equipment) link 2 CE A1850 TSS-320 (PE) 1626LM Application Client / Server link 3 Univ. Orsay (PoP 2) link 4 Transmissions: λ: 4x10 Gb/s 40 Gb/s Network Controller PE 1626LM CEA Ter@tec (PoP 4) CE Application Client / Server 10Gb/s Interfaces (SDH or Ethernet: aggregation over λ) 31

32 1354BM 1353NM Network Management Issy les Moulineaux Local monitoring CARRIOCAS Pilot Network and test bed configuration 4 x 10GEth. Clamart Picture wall Transport Service Switch Carrier Grade Ethernet with T-MPLS capabilities 1850 TSS 4 x 10GEth. Trib 40G prototypes fournis gracieusement par Alcatel dans SP1 durant la vie du projet Local monitoring 4 x 10GEth. storage 4x10 GEth. Access Saclay Légende Orsay storage 40 Gbps Transmission 1626 LM Super-computers Ter@tec 4 x 10GEth. Bruyères Le Châtel Lambda 1 à 10G omnibus Lambda 2 à 10G Lambda 3 à 10G Lambda 4 à 10G Lambda 5 à 40G Nx10 Gbps Transmission Deployment phase 1 (10Gb/s): June 2007 Deployment phase 2 (40Gb/s): July

33 Contents 1. CARRIOCAS and Motivations for Utility Computing in Business 2. Distributed Application classification function of their requirements 3. Network Services definition in Open Service Architecture 4. Efficient Network architecture combining capacity and flexibility 5. Grid Application and Network interactions: Needs and Levels 6. Scheduling, Reconfiguration & Virtualization (SRV) service 7. CARRIOCAS Pilot Networks 8. Conclusion 33

34 CARRIOCAS: path towards Industrial Grid computing over Internet Prepare ultra-high bit rate transport networks Optimal bit rate transmission / cost ratio Management/control interactions between applications and networks Taking into account business companies requirements Expected CARRIOCAS next-steps Enlarge the offer of computing intensive applications Attract new users: virtual organizations Extend the network infrastructure coverage 34

35 Many Thanks to: CARRIOCAS partners and colleagues the French Ministry of Industry, Essonnes, Haut-de-Seines and Paris General Council for their supports on CARRIOCAS project FEDERICA-PHOSPHORUS projects at TERENA Networking Conference