1 / 29

Pete playing devils advocate

Pete playing devils advocate. JRA4-F2F. CERN 14/15 OCt 200 4. www.eu-egee.org. EGEE is a project funded by the European Union under contract IST-2003-508833. Javier Orellana EGEE-JRA4 Coordinator. Critique of Subactivity: NPM. What are we trying to achieve ? See TA high level description.

etina
Télécharger la présentation

Pete playing devils advocate

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Pete playing devils advocate JRA4-F2F CERN 14/15 OCt 2004 www.eu-egee.org EGEE is a project funded by the European Union under contract IST-2003-508833 Javier OrellanaEGEE-JRA4 Coordinator

  2. Critique of Subactivity: NPM • What are we trying to achieve ? • See TA high level description <BAR Survey, Cambridge> <8 September 2004> - 2

  3. Critique of Subactivity: NPM • What is it primarily about • Standardising meaning of NP information • Deploying standardised access to NP information for a range of consumers • First opportunity for coherent scheme including EGEE sites and GNs domains • Recasting existing work into web services framework • Getting agreement form all sites and domains to deploy a standard framework (regardless of its initial content) • Developing diagnostic tools for GOC and ROC • Making the information useful to Higher Layer Middleware (HLM) in EGEE • What is it NOT about • Developing yet another implementation with no context (enough of these exist: WP7, IEPM, Pipes, GridMon, Perfmonit, MonaLisa). These are all potential backends/frontends. • Deploying an arbitrary implementation at EGEE and other sites (we could have just used WP7 for this). • Monitoring tool development (IPerf, Ping….) (this is secondary) • Worrying unduly about what measurements mean (at least in first instance) <BAR Survey, Cambridge> <8 September 2004> - 3

  4. Strawman architecture (not rpoperly designed yet) EGEE Resource Information System JRA4 code to publish (& interpolate ?) Interface based on GGF-NMWG Imp 3 Imp 3 Imp 1 NREN Geant Inet-2 <BAR Survey, Cambridge> <8 September 2004> - 4

  5. Strawman architecture (not rpoperly designed yet) GRM JRA4 code to query for GRM Interface based on GGF-NMWG Imp 3 Imp 3 Imp 1 NREN Geant Inet-2 <BAR Survey, Cambridge> <8 September 2004> - 5

  6. Strawman architecture (not rpoperly designed yet) Diagnostic Services for GOCs JRA4 code For network diagnostics Interface based on GGF-NMWG Imp 3 Imp 3 Imp 1 NREN Geant Inet-2 <BAR Survey, Cambridge> <8 September 2004> - 6

  7. Where might existing work fit in this ? EGEE Resource Information System WP7 R-GMA work ?? Perfmonit Imp 3 WP7 NREN Geant Inet-2 <BAR Survey, Cambridge> <8 September 2004> - 7

  8. GN2 code Interface based on GGF-NMWG Imp 3 Imp 3 Imp 1 (WP7?) NREN Geant Inet-2 <BAR Survey, Cambridge> <8 September 2004> - 8

  9. Critique of Subactivity: NPM • How to plan • Develop a project plan based upon what we need to achieve by when • Look at bigger picture than simply the “ titles of the deliverables” • Realise this is first opportunity for coherent scheme including EGEE sites and GNs domains, and just getting a basic framework deployed will be a major achievement. Getting the content perfect is a secondary goal. • Re assign effort to achieve (i) consolidation i.e. no 0.5 PM here and there (ii) use strengths • Put in place a much more formal project mangement structure so that every one knows what they are meant to do • Milestones and deliverables • Merely a punctuation of project plan – not a definitive receipe for success • These DO NOT define a complete project plan. • Current WBS • Only a first iteration • Fine up to PM6 in most places (requirements, surveys) • Inadequate in terms of getting a framework deployed in a timely fashion • Not enough understanding of time needed to get first iteration of architecture & interfaces defined • No explicit planning for a long and resource intensive software development cycle to get prototype framework on ground by PM9 (this is a fault of the TA) <BAR Survey, Cambridge> <8 September 2004> - 9

  10. Critique of Subactivity: NPM • Key omissions in WBS • Understanding that prototype work is needed to define interfaces • You don’t just write them down and assume they work • Understanding that web services framework development is needed for any of this to mean anything, and that this needs • a proper canonical software development approach, • a long lead time • This has nothing to do with measurement implementation, which is a back end to the framework • Should have started much earlier • Understanding that defining an AA solution requires serious effort <BAR Survey, Cambridge> <8 September 2004> - 10

  11. Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 In my opinion Requirements… Target: Something working which is a step towards overall goals by ~ PM10 Why: WP7 exists NMWG exists Why take longer ? Surveys… Prototype WS based standard interface framework deployed Arch 1 Interface 1 AA 1 WS-Framework 1 Arch 2 Interface 2 AA 2 WS-Framework 2 Backend for EGEE sites Simple Client <BAR Survey, Cambridge> <8 September 2004> - 11

  12. Performance Monitoring <BAR Survey, Cambridge> <8 September 2004> - 12

  13. Bandwidth on Demand <BAR Survey, Cambridge> <8 September 2004> - 13

  14. Configure Site Broker Net Broker Configure Configure Net Broker Net Broker Configure Configure Site Broker Geant NREN-B NREN-A Site Site Network Resource Allocation and Reservation Source Dest EGEE Middelware EGEE Middelware <BAR Survey, Cambridge> <8 September 2004> - 14

  15. Geant NREN-B NREN-A Site Site Static SLA Static SLA Static SLA Static SLA DJRA4.1 EGEE Middleware Resource Provider EGEE-Network Liaison AAA Configure Configure [Note: For example, GRS project of S.Bhatti, UCL] <BAR Survey, Cambridge> <8 September 2004> - 15

  16. IPv6 <BAR Survey, Cambridge> <8 September 2004> - 16

  17. JRA4 IPv6 Policy Statement • EGEE agreed with 6NET at the time of the original proposal to work with 6NET to • promote IPv6, • make EGEE software developers aware of IPv6 coding practice, • investigate the possibility of trying some EGEE code on a limited IPv6 testbed • This agreement is codified in the relevant deliverable DJRA4.3 • This has not been critical path in comparison to initial requirements gathering deliverables an milestones, and hence it was natural and timely to leave this until after PM6 • JRA4 has now started on this. We have • held a preliminary meeting with Piers O'Hanlon from 6NET • spoken at high level with senior 6NET personnel (Kirstein, Butler) to re-affirm intentions • We remain as committed as ever to work with 6NET, and our intentions are to • hold sessions on IPv6 awareness raising (NeSC training) • promote good practice for writing code independent of IPv4/6 (NeSC training • Investigate whether some joint IPv6 testbed work is feasible. • We now plan to hold a telecon soon in October to take this forward • I <BAR Survey, Cambridge> <8 September 2004> - 17

  18. Middleware service Site 1 Site 2 Domain 3 Domain 1 Domain 2 <BAR Survey, Cambridge> <8 September 2004> - 18

  19. Data Volume V transferred from A to B in time T Min/Max bandwidth over time T from a to B IP Premium width DSCP C in this domain for bandwidth B in a period T MPLS-VPN tunnel in this domain for bandwidth B in a period T Domain B Domain C Domain A Suggestion 1: Hierarchical Middleware EGEE Grid Middleware Transport Service Network Element LOGICAL User Service Network oriented service requestor Site Site LOGICAL Service LOGICAL Service LOGICAL Service Physical Service Physical Service Physical Service <BAR Survey, Cambridge> <8 September 2004> - 19

  20. Domain C Domain B Domain A Suggestion 2: Sequential Middleware EGEE Grid Middleware Transport Service Network Element LOGICAL Service Network Oriented Service requestor Site Site LOGICAL Service LOGICAL Service LOGICAL Service Physical Service Physical Service Physical Service <BAR Survey, Cambridge> <8 September 2004> - 20

  21. NM-WG NM-WG NM-WG SP Tool X Tool Y Tool Z R-GMA Middleware Domain A Domain B Domain C Network Performance Architecture <BAR Survey, Cambridge> <8 September 2004> - 21

  22. Interfaces with Grid Middleware Authentication Authorisation Auditing Grid Access Service Information and Monitoring Network Performance Monitoring Grid Middleware Accounting Data Management Site Gatekeeper Workload Management Bandwidth Allocation and Reservation <BAR Survey, Cambridge> <8 September 2004> - 22

  23. Review WBS [P.Clarke] • Review and assign effort allocation to this task, a sit needs both • Network experts • Experienced software developers <BAR Survey, Cambridge> <8 September 2004> - 23

  24. Components and Requirements Consumer (User Application) Operations Interface End-user Interface Grid Middleware Operations Middleware Interface Computer Element Network Element Storage Element NE-A NE-B NE-X Network (GEANT+NRENs) <BAR Survey, Cambridge> <8 September 2004> - 24

  25. 5.- Resource Management 4.- Path Discovery 3.- Can Allocate Resource ? 2.- Topology Discovery 5.- Resource Management 1.- Authentication Authorization C A B Resource Management Path Discovery END Can Allocate Resource ? Topology Discovery NOW !!! Authentication Authorization BB-B BB BB-A BB-C BB Bandwidth Broker Traffic flows Signalling between BB Provisioning devices Bandwidth Allocation & Reservation 4.- Path Discovery 3.- Can Allocate Resource ? 2.- Topology Discovery NREN C 1.- Authentication Authorization NREN A GEANT NREN B ie.: One flow from A to B <BAR Survey, Cambridge> <8 September 2004> - 25

  26. 2.- Path Discovery 1.- Authentication Authorization NOW !!! A B C 4.- Resource Management 3.- Can Allocate Resource ? 2.- Topology Discovery 1.- Authentication Authorization BB-C MultiDomain-BB BB-B BB BB-A BB Bandwidth Broker Traffic flows Signalling between BB Provisioning devices Bandwidth Allocation & Reservation NREN C NREN A GEANT NREN B ie.: One flow from A to B <BAR Survey, Cambridge> <8 September 2004> - 26

  27. Grid Middleware Grid Users Network Performance Monitoring • Grid Performance closely linked to Network Performance • Network Performance?, what for? : • Problem diagnostic and rectification • Facilitate resources allocation • Performance monitoring and SLA adherence GOCs NOCs Operations NOC: Network Operation Center GOC: Grid Operation Center Information Service (R-GMA) Network (GEANT+NRENs) Common Interface (OGSA) Domain_A Domain_B . . . . Domain_X <BAR Survey, Cambridge> <8 September 2004> - 27

  28. A B PMS 3 PMS 2 PMS 1 DM 3 DM 1 DM 2 DM Domain Manager PMS Performance Monitoring System Signalling between DM Request of Measurement Net. Perf. Monitoring: Use case example ie. OWD from point A to B ? • PMSx and DMx • Are independent implementation for the measurements • Features • Multiple domains, AAA, OGSA/OGSI We DON’T yet We got that already NREN A GEANT NREN B OWD=OWD1+OWD2+OWD3 <BAR Survey, Cambridge> <8 September 2004> - 28

  29. Thank you Questions? <BAR Survey, Cambridge> <8 September 2004> - 29

More Related