Federated E-infrastructure Dedicated to European Researchers Innovating in Computing network Architectures

Transcription

1 Federated E-infrastructure Dedicated to European Researchers Innovating in Computing network Architectures Mauro Campanella - GARR On behalf of the FEDERICA project CONVEGNO "Incontro con l'ingegneria delle Telecomunicazioni" Firenze, June 18th, 2008

2 Agenda FEDERICA at a glance, vision and principles Description and status Boundaries and (control plane) challenges 2

3 FEDERICA at a glance What: When: Who: European Community co-funded project in its 7 th Framework Program in the area Capacities - Research Infrastructures 3.7 MEuro EC contribution, 5.2 ME budget, 461 Man Months 1 st January June 2010 (30 months) 20 partners, based on stakeholders on network research and management: 11 National Research and Education Networks, DANTE (GÉANT2), TERENA, 4 Universities, Juniper Networks, 1 small enterprise (MARTEL), 1 research centre (i2cat) - Coordinator: GARR (Italian NREN) Where: Europe-wide e-infrastructure, open to external connections 3

4 FEDERICA Vision Create by fall 2008 an e-infrastructure for all researchers on Future Internet. Allow researchers complete control of resources in a slice, enabling disruptive experiments. Support research in virtualization of e-infrastructures integrating network resources and nodes capable of virtualization (V-Nodes). In particular on multi(virtual)domain control, management and monitoring, including virtualization services and user oriented control in a federated environment Pave the way, research and create experience for the next generation of the European Research and Education Networks. 4

5 Partners Location JUNIPER Networks NORDUnet KTH NREN Univ. or Res. Centre SME, Associations, Vendors HEAnet DANTE TERENA PSNC NREN partners provide a European coverage using the GN2+ service and FCCN Red.ES i2cat UPC Dark Fiber GÉANT2 GÉANT+ service Cross Border Fiber or future GÉANT2 Martel Switch PoliTO GARR DFN CESNET 5 NIIF Hungarnet ICCS GRNET - allow connection to Univ. and Research Center partners - Provide HUB functionalities and possibility extend the e- Infrastructure to other countries and projects using physical or logical circuits - Contribute with tools and specific expertise

6 FEDERICA Partners National Research & Education Networks (11) CESNET Czech Rep. DFN Germany FCCN Portugal GARR (coordinator) Italy GRNET Greece HEAnet Ireland NIIF/HUNGARNET Hungary NORDUnet Nordic countries PSNC Poland Red.es Spain SWITCH Switzerland Small Enterprise Martel Consulting Switzerland NREN Organizations TERENA DANTE The Netherlands United Kingdom Universities - Research Centers i2cat KTH ICCS (NTUA) UPC PoliTO System Vendor Juniper Networks Spain Sweden Greece Spain Italy Ireland 6

7 The Enabling Elements 1. Virtualization in computing systems and in network is available. It creates resources, given a supporting physical substrate, which : - Have a looser or no dependency from a specific physical location or entity (computing, data, circuits may migrate) - On-the-fly reconfiguration, cancellation and creation of resources in the e-infrastructure (e.g. a routing element, a generic resource) - off-the-shelf components offer in hardware, embedded virtualization functionalities. 2. The European NRENs are managing owned hybrid infrastructures and actively performing network research, starting from users needs. The federated NREN architecture scenario offers now significant interdomain services and research capabilities. 3. The traditional testbed, focused on a small number of technologies has a usefulness limited to the specialized nature of users. It also implies a long set-up time and a fast obsolescence. 7

8 GN2 project The GÉANT2 network European NRENs Backbone based on dark fibers operated by DWDM at multiple 10 Gb/s 8

9 The Network GARR-G Connects all Universities and research centres in Italy, plus ~1000 schools, Libraries, Research Hospitals (50 IRCCS), Backbone aggregated capacity: ~120 Gbps Access aggregated capacity: ~60 Gbps 615 user access circuits from 2Mbps to 10 Gbps 62 backbone links 43 POP (90% hosted by users) PEERING: 65 Gbps 42.5 Gbps to GEANT2 3 x 2.5 Gbps IP Transit 5 x 1 Gbps+10 Gbps National Circuits leased from: 8 national operators (Telecom Italia, Infracom, Fastweb, Interoute, WIND BT-Italy, COLT) 3 international operators (Global Crossing,Telia, Level3) 9

10 FEDERICA e-infrastructure 10

11 FEDERICA Principles 1. Create an agnostic and neutral (transparent) infrastructure 2. Create slices which are a set of (virtual) network and computing resources according to user s request and are disruptible 3. Provide to the user complete control within a slice down to the lowest possible layer (in particular allow any application and protocol) 4. Strive/engineer for reproducibility of experiments, i.e. given the same initial conditions, the results of an experiment are the same 5. Allow slices (if requested) to connect to the general Internet, to access external services/nodes (e.g. for content/delivery, specialized HW) 6. Ensure isolation between slices maintaining the possibility to crossconnect slides on request 7. Allow simultaneous use without conflict 8. Force/be exposed to topology changes (various level of resiliency) 9. Open to interconnect / federate with other e-infrastructures 10. Access granted through a User Policy Board 11

12 Agenda FEDERICA at a glance, vision and principles Description and status Boundaries and (control plane) challenges 12

13 Work plan outline (I3) Jan 2008 Oct 2008 Feb 2010 Slices 13

14 FEDERICA Activities NA1: Project Management NA2: Building and Consolidating the User Community NA3: Standardization and Liaisons NA4: Dissemination and Training Network Activities SA1: Infrastructure Support SA2: Operational User support and Tool bench development Service Activities JRA1: Network Control and Management JRA2: Novel Paradigms and User Control Join Research Activities 14

15 Pictorial of Slices Creation The user requests an Infrastructure made of L2 circuits, un-configured virtual nodes, to test a new BGP version. NRENs and Global Internet Slice 1 FEDERICA substrate FEDERICA Network Layer substrate FEDERICA Data Link layer substrate FEDERICA Physical Layer substrate Virtual Router/Switch Virtual node 15

16 FEDERICA e-infrastructure DRAFT 23-Apr-08 Core circuits and switches being delivered Core Nodes 16

17 Topology version 8.4 NORDUNET SUNET KTH HEAnet DFN PSNC SWITCH GARR CESNET Hungarnet FCCN Red.es GRNET i2cat Legenda Core Nodes 1 Physical GbE from GN2+ 1 GbE VLAN or L2MPLS 1 Physical GbE tbd A high physical meshing allows a better distribution of slices in the substrate. Long delay and high capacity circuits are also available. 17

18 Sample FEDERICA PoP Out-Of-Band Terminal server Other FEDERICA PoPs Notes: - Each PC has many Gb Ethernet interfaces - The FastEthernet Interfaces are to decouple the control and data plane - OOB is not mandatory FEDERICA Computing Node (PC) FEDERICA Network switch BGP Peering NREN Production Network Management Legenda GigabitEthernet FastEthernet RS-232 The FEDERICA substrate (physical infrastructure and Single IP AS public number) 18

19 Slicing the Core (Substrate) FEDERICA substrate Switches: Juniper MX480, (virtual and logical routing, MPLS, VLANs, IPv4 v6, QoS linecards) V-Nodes: Up to 8-16 images/node, Unix OS, 4-8 Ethernet NICs, ~ 1 TB disk, 4core CPUs 19

20 Access and Use Policies A User Policy Board will receive and approve project for the use of the infrastructure Access to the infrastructure is subject to the signature of an Acceptable User Policy, which includes providing feedback Usage of a FEDERICA slice will be free of charge if no additional equipment is requested Interconnection with other infrastructures (e.g. labs) is possible. The cost is to be defined case by case. The time duration of a single experiment will be in principle limited to facilitate turnover (3 months initially) Access is open to research groups from academia and private sector with priority to European Community funded projects. The code and tool bench produced will be Open Source Users requirements are fundamental and are being collected 20

21 Federating and Interconnection with other projects/groups Soft model Act as a user of the project (hosting its virtual resources with possible interconnection to external resources using Internet) Collaboration at the project/user group level to exchange intermediate results, perform joint developments Hard model Hosting in FEDERICA a physical resource Interconnecting the testbeds using a dedicated network resource Need for a resource/service description format (e.g. TMF, IPSphere, ) 21

22 FEDERICA - Goals Summary In scope Provide on European scale network and system agnostic e-infrastructure to be deployed in phases for Future Internet research (and not only). Provide its operation, maintenance and on-demand configuration Act as a forum and support for researchers/projects on Future Internet. Support of experimental activities to validate theoretical concepts, scenarios, architectures, control and management solutions. Users have full control of their slice Validate and gather experimental information for the next generation of research networking also through basic tool validation Dissemination and cooperation between NRENs and researchers community Contribution to standards in form of requirements and experience Out of scope Internal extended research, e.g. advanced optical technology Development of Grid applications (but open to hosting) Offer raw computing power Offer transit capacity 22

23 Agenda FEDERICA at a glance, vision and principles Description and status Boundaries and (control plane) challenges 23

24 Initial Boundaries Scalability IPv6 unconfigured, to be enabled according to users requests Ethernet framing (large MTUs) as data link Packet switching and statistical multiplexing assumed by default Less powerful switches outside the core Initial manual provisioning Not all technologies available (e.g. wireless, nomadic nodes) Larger virtual slices can be obtained reducing the number of concurrent users, user s equip. may be added Equipment is ready for IPv6 Not considered a limiting factor, can be overcame later using WDM equipment Hardware QoS is available on two Juniper MX480 switches Rely more on software emulation Slower initial provisioning, compatible with decision process and overall management Equipment hosting, federation with other testbed 24

25 Control Plane and Virtualization: an additional challenge - The main control plane is now tied to the basic substrate, its main and new function is to create slices made of resources. It implies the need to define: - user interface specification for request (UNI-type, but UNIs do not have currently the required parameters) - resources abstraction and specification (network part ongoing in GN2 (AutoBAHN) and international collaboration with I2, ESnet) - signalling protocols for UNI communication - Interdomain communication specification - Need to manage the basic substrate and one or more slices at the same time: - Control plane extensions in the layered framework - control plane capable of communication with one or more hosted control planes 25

26 C.P. and Virtualization (cont) - Extension and definitions for naming, addressing, to allow the possibility of global reachability - Monitoring interaction,synergy between the substrate and the slices in sharing the physical elements FEDERICA will start with manual provisioning of virtual resources, a single IP AS and public IP network addresses for the substrate. Overall neither the integrated model, nor the layered one may be adequate and further advances are required in HW and SW to allow virtualization capabilities with resource assurances. BUT 26

27 THE challenge: Complexity Internet as BGP Peerings (CAIDA) Internet as Circuits? Virus Sobig-F ( F-secure) The major risk is to create a system which is too complex. 27

28 Grazie per l attenzione 28

29 FEDERICA Partners 29

30 GARR L acronimo GARR deriva da Gruppo per l Armonizzazione delle Reti della Ricerca. Il Gruppo si e autocostituito nel 1987 per permettere una sinergia delle comunità scientifica delle Università e degli Enti Pubblici di Ricerca italiani nella realizzazione, gestione, ricerca e utilizzazione delle reti di interconnesione, analogamente alle Reti della Ricerca e dell Istruzione già esistenti in altri paesi. Nel 1989 Il GARR si costituisce formalmente grazie ad un progetto finalizzato del MURST Nel 2002 nasce l Associazione senza fini di lucro Consortium GARR, essendo CNR, CRUI, ENEA ed INFN i membri fondatori Il mandato principale del Consortium GARR e di gestire e sviluppare la rete della ricerca e dell istruzione italiana. 30

31 La storia del GARR GARR è uno strumento per contribuire alla armonizzazione e sinergia tra mondo dell istruzione e della ricerca 1987: Inizio collaborazione per le reti fra enti di ricerca ed Università 1989: GARR fondato con un progetto ministeriale (MURST) come gruppo di lavoro per l armonizzazione delle reti della ricerca : GARR-1, prima rete comune a 2Mbps : GARR-2, evoluzione di GARR-1 a 34Mbps : GARR-B (Broadband) a 155Mbps 2002: il GARR diventa Consortium GARR un un entità legale senza fini di lucro (fondatori CNR, ENEA, INFN, CRUI) : sviluppo di GARR-G (Giganet) a Gbps 2008 GARR-X 31

32 Gigabit/sec Fase POP: 15 TI + 1 GARR 1 Carrier BackBone 1Gbps GARR-B Fase Fase POP: 15 TI + 4 GARR 1 Carrier Fase BackBone 54Gbps GARR-G Fase1 Fase POP: 7 TI + 30 GARR 8 Carrier Accesso 29Gbps Fase Gigabit/sec 10 Accesso 500Mbps