![10
Paths from WLCG Tier1s
Most of the sites would send data through their Commercial Internet
upstream (or those of their NRENs).
Only few NRENs had peerings with T-Systems at IXPs
CA-TRIUMF-> BCnet, HE, DECIX, T-Systems
CH-CERN-> GEANT-VRF, DFN, T-Systems
DE-KIT-> Belwue, T-Systems
ES-PIC-> CESCA, REDIRIS, NTT, Netuse, T-Systems
FR-CCIN2P3-> Renater, C&W, DTAG, T-Systems [from Renater LG]
IT-INFN-CNAF-> Cogent, NTT, Netuse, T-Systems
NDGF-> NORDUnet, DECIX, Global-Connect, Netuse, T-Systems [from NORDUnet LG]
NL-T1-> SURFnet, NORDUnet, DECIX, Global-Connect, Netuse, T-Systems
RRC-KI-T1-> RETN, Netuse, T-Systems
TW-ASGC-> Chunghwa, Sprint, DTAG, T-Systems
UK-T1-RAL-> Janet, Telia, Netuse, T-Systems
US-FNAL-CMS-> ESnet, DTAG, T-Systems
US-T1-BNL-> ESnet, DTAG, T-Systems](https://image.slidesharecdn.com/tnc17slidestnc17-cern-and-public-clouds-180314121926/85/Network-experiences-with-Public-Cloud-Services-TNC2017-10-320.jpg)
Network experiences with Public Cloud Services @ TNC2017
- 1. Network experiences with Public Cloud Services TNC2017 30th of May 2017 edoardo.martelli@cern.ch
- 2. 2 Content - CERN experiences with Public Clouds - HNSciCloud - RENs’ connectivity solutions
- 3. 3 CERN Rationale for Public Clouds - Additional, temporary resources to fit burst of computing demand - Potentially replacing on-premise capacity Potential benefits: - Economy of scale - Elastic, adapts to changing demands - Delegation of hardware procurement and operations Potential challenges: - Integration with on-premise resources (orchestration) - Compute-heavy tasks relatively new to cloud market (doesn’t fit overbooking) - Cloud provider’s business models vs. procurement rules and procedures of public organizations
- 4. CERN past experiences with Public Clouds
- 5. 5 CERN approach to procurement Series of short procurement projects of increasing size and complexity
- 6. 6 Last procurement (2016) CERN awarded a contract to T-Systems for resources to be used for physics data processing Size of procured resources for 3 months (+ 1 month of data extraction) - 1’000 simultaneous VMs (4’000 cores), - RAM: 8 GB/VM, Local Disk 100 GB/VM - 500 TB of Central Storage - 10Gbps connectivity to GEANT - 1’000 public IPv4 addresses (resources accessible from all WLCG)
- 7. 7 WAN connectivity DT AS3320 OTC DFN GEANT Cloud VRF AS20965 A C Three options available: a) Peering with GEANT Cloud VRF through DFN b) Existing 7Gbps Internet upstream c) Existing alternative path via GEANT IAS and NORDUnet TSI AS34086 GEANT IAS B NORDUnet SIOL Global Connect CERN AS513
- 8. 8 Performance over GEANT Cloud VRF On a day On a month
- 9. 9 Network throughput vs number of jobs 30-Aug-2016 9
- 10. 10 Paths from WLCG Tier1s Most of the sites would send data through their Commercial Internet upstream (or those of their NRENs). Only few NRENs had peerings with T-Systems at IXPs CA-TRIUMF-> BCnet, HE, DECIX, T-Systems CH-CERN-> GEANT-VRF, DFN, T-Systems DE-KIT-> Belwue, T-Systems ES-PIC-> CESCA, REDIRIS, NTT, Netuse, T-Systems FR-CCIN2P3-> Renater, C&W, DTAG, T-Systems [from Renater LG] IT-INFN-CNAF-> Cogent, NTT, Netuse, T-Systems NDGF-> NORDUnet, DECIX, Global-Connect, Netuse, T-Systems [from NORDUnet LG] NL-T1-> SURFnet, NORDUnet, DECIX, Global-Connect, Netuse, T-Systems RRC-KI-T1-> RETN, Netuse, T-Systems TW-ASGC-> Chunghwa, Sprint, DTAG, T-Systems UK-T1-RAL-> Janet, Telia, Netuse, T-Systems US-FNAL-CMS-> ESnet, DTAG, T-Systems US-T1-BNL-> ESnet, DTAG, T-Systems
- 11. Helix Nebula Science Cloud
- 12. 12 HELIX NEBULA the Science Cloud Procurers: CERN, CNRS, DESY, EMBL-EBI, ESRF, IFAE, INFN, KIT, SURFSara, STFC - Procurers have committed funds (>1.6M€), manpower, use- cases with applications & data, in-house IT resources Objective: procure innovative IaaS level cloud services - Fully and seamlessly integrating commercial cloud (Iaas) resources with in-house resources and European e- Infrastructures - To form a hybrid cloud platform for science Services will be made available to end-users from many research communities: High-energy physics, astronomy, life sciences, neutron/photon sciences, long tail of science Co-funded via H2020 (Jan’16-Jun’18) as a Pre-Commercial Procurement (PCP) project: Grant Agreement 687614, total procurement volume: >5M€
- 13. 13 HNSciCloud challenges Innovative IaaS level cloud services integrated with procurers in-house resources and public e-infrastructure to support a range of scientific workloads Compute and Storage support a range of virtual machine and container configurations including HPC working with datasets in the petabyte range Network Connectivity and Federated Identity Management provide high-end network capacity via GEANT for the whole platform with common identity and access management Service Payment Models explore a range of purchasing options to determine those most appropriate for the scientific application workloads to be deployed
- 14. 14 High Level Architecture of the Hybrid Cloud Platform including the R&D challenges
- 15. 15 Network Requirements (LAN and WAN) - Storage nodes globally accessible via public IP addresses (1:1 NAT accepted) - Computing nodes can reside behind NAT - At least 1Gbps capacity between pairs of local nodes - At least 10Gbps of aggregated capacity between Supplier and Customer nodes - WAN connectivity via one of the following - a. an NREN with transit to the customer; - b. GEANT or NORDUnet; - c. an IXP which the customer is also connected to - Support of identity federation (eduGAIN) for IT managers
- 16. 16 Project phases We are here
- 17. 17 Selected bidders Admitted to the Design phase: - T-Systems, Huawei, Cyfronet, Divia - IBM - RHEA Group, T-Systems, exoscale, SixSq - Indra, HPE, Advania, SixSq Other major players not interested or dropped out just before tender submission Admitted to the Prototype phase: - T-Systems, Huawei, Cyfronet, Divia - IBM - RHEA Group, T-Systems, exoscale, SixSq Current Status: Evaluation of bids in the Prototype Phase
- 18. 18 Prototype phase Collect a number of tests from the Procurers - to test/validate the design proposed for the several project challenges - to be able to assess results of the tests as one of the main criteria to be eligible to the next phase (Pilot) Tests consist in applications and some selected workloads Next phase (Pilot) will consist in the assessment of the prototypes proposed on: - performance - scalability - security
- 20. 20 R&E connectivity to Public Clouds Challenges - Bridge R&N and Commercial Networks with enough bandwidth - Reduce cost of data transfers - RENs policies limiting transport of “commercial” traffic (i.e. how to move R&E data between two cloud storage centres )
- 21. 21 GEANT: dedicated VRF GEANT GEANT CLOUD VRF GEANT NREN (AS200) PEERING REN (AS300) IX GEANT NREN (AS100) BGP announcements No tag NREN tag Cloud Provider tag deny AS11 CP2 (AS22) CP3 (AS33) 64512:11 AS22, AS33 CP1 (AS11) AS200,AS300 Slide credit: GEANT
- 22. 22 ESnet: on-net VPN termination Virtual “Site Router” (VRTR) Service At the edge of the cloud Virtual “Site Router” improves path efficiency and takes pressure off of the site local-loop. Slide credit: ESnet
- 23. 23 NORDUnet: transport to eXchange Points Slide credit: NORDUnet
- 24. 24 Best practices for CSP connectivity NORDUnet presented a document which suggest a way of connecting CSPs and their customers by using Open eXchange Points and connectivity provided by the REN The document aims to simplify the procurement of the connectivity to CSPs
- 25. 25 Network connectivity options for CSPs CERN IT-CS is writing a document which compare the different connectivity options proposed by the RENs Document https://indico.cern.ch/event/527372/contributions/2236895/subcontributions/208050/attachments/1338702/2015050/connectivity-options-for-clouds-draft.pdf
- 26. Conclusions
- 27. 27 Summary CERN had positive preliminary tests with Commercial Cloud Providers. Ad-hoc connectivity through GEANT gave much better throughput HNSciCloud has reached the Prototype phase. Parties are progressing on connectivity through GEANT and NRENs RENs are working with institutes and Cloud providers to best fit connectivity needs