deep h hybrid rid datacloud ud
play

DEEP-H -Hybrid rid-DataCloud ud Digi gita tal I Inf nfrastr - PowerPoint PPT Presentation

DEEP-H -Hybrid rid-DataCloud ud Digi gita tal I Inf nfrastr tructu uctures f for R Research ( ch (DI4R) lva varo L Lpez Garc ca Lisbon (Portugal) aloga@ifca.unican.es October 10, 2018 Spanish National Research Council


  1. DEEP-H -Hybrid rid-DataCloud ud Digi gita tal I Inf nfrastr tructu uctures f for R Research ( ch (DI4R) Álva varo L López Garcí cía Lisbon (Portugal) aloga@ifca.unican.es October 10, 2018 Spanish National Research Council DEEP-HybridDataCloud has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 777435.

  2. Our Our vis visio ion ● We need to build added value and advanced services on top of bare IaaS and PaaS infrastructures ● Ease and lower the entry barrier for non-skilled scientjsts – Transparent executjon on e-Infrastructures – Build ready to use modules and ofger them through a catalog or marketplace – Implement common sofuware development techniques also for scientjst’s applicatjons (DevOps) ● Build and promote the use of intensive computjng services by difgerent research communitjes and areas, an the support by the corresponding e- Infrastructure providers and open source projects DEEP-HybridDataCloud 10/10/18 2/14

  3. Who ho is is the us user er? machine chine le learni ning ng expert rtise t t level of knowledge being e e c c domain expertise h h required depends on the n n o o specifjc use case and the user profjle l l o o g g i i c c a a l l e e x x p p e e r r t t i i s s e e DEEP-HybridDataCloud 10/10/18 3/14

  4. Deep eep lea earnin ning us use e ca cases es ● Category 1: Deploy a readily trained network for somebody else to use it on his/her data set ● Domain knowledge ● Category 2: Retrain (parts of) a trained network to make use of its inherent knowledge and to solve a new learning task ● Domain + machine learning knowledge ● Category 3: Completely work through the deep learning cycle with data selectjon, model architecture, training and testjng ● Domain + machine + technological knowledge DEEP-HybridDataCloud 10/10/18 4/14

  5. Previo evious usly ly... ● Scientjsts create a deep learning applicatjon on their personal computers ● The deep learning model is trained in a GPU node (maybe also locally) – What happens if they do not have access to one? ● The work is published (or not) – Model architecture, confjguratjon, scientjfjc publicatjon, etc. ● But: – How can a scientjst easily ofger it to a broader audience? – What about dependencies? DEEP-HybridDataCloud 10/10/18 5/14

  6. Ofg Ofger ering ng develo developed ed mo models els as a ser ervice ● Development of APIs and web applicatjons ● Scientjsts need to know what an API is – REST, GET, POST, PUT... ● Lack of API consistency → hard for external developers to consume them ● Provide users with a generic API (OpenAPI) component where they applicatjon can be plugged DEEP-HybridDataCloud 10/10/18 6/14

  7. Ser Service ice compo mposit itio ion ● An applicatjon may consist on several components that need to be deployed, confjgured, etc → service compositjon ● Service compositjon, if done properly, provide a way to re-deploy the same topology over difgerent infrastructures → catalog of components ● Scientjsts should not need to deal with technologies and infrastructures they do not care at all ● We need therefore difgerent roles, to perform difgerent tasks – Comparison with laboratory technician ● In INDIGO-DataCloud we started with this approach, but this needs to be generalized (and the roles recognized) DEEP-HybridDataCloud 10/10/18 7/14

  8. DEEP P Open Open Cat atalo log ● Collectjon of ready-to-use modules – Comprising machine learning, deep learning, big data analytjcs tools – ML Marketplace htups://marketplace.deep-hybrid-datacloud.eu – GitHub � htups://github.com/indigo-dc?utg8= &q=DEEP-OC – DockerHub htups://hub.docker.com/u/deephdc/ ● Based on DEEPaaS API component – Expose underlying model functjonality with a common API – Based on OpenAPI specifjcatjons – Minimal modifjcatjons to user applicatjons. ● Goal: execute the same module on any infrastructure: DEEP-HybridDataCloud 10/10/18 8/14

  9. Difg ifger erent ent roles les fo for difg difgerent ent tas asks U s e r I n p u t S a a S U s e r A c c e s s P o r t a l O u t p u t App U t i l s S o l v e r P a a S E n a b l e r A R e p o s i t o r i e s O r c h e s t r a t o r A I R e s o u r c e R e s o u r c e P r o v i d e r P r o v i d e r e - I n f r a s t r u c t u r e s ( F e d e r a t i o n s ) D o c k e r ( a p p ) D o c k e r ( u t i l s ) A P I s V M S i t e A S i t e B C o m p u t e C o m p u t e N e t w o r k S t o r a g e I a a S 1 : N DEEP-HybridDataCloud 9/14

  10. DEEP P hig high h lev evel l Archi hitecture DEEP-HybridDataCloud 10/10/18 10/14

  11. P 1 st rele DEEP elease: e: co comp mponen nents Softw tware component Functionalities Hybrid deployments on multiple sites ● PaaS Orchestrator Support to specifying specialized computing hardware ● Improved support for deployment failures ● Improved support for hybrid deployments ● Infrastructure Manager (IM) Support for additional TOSCA types ● Support for training a machine learning application ● Support for performing inferences/analisys/predictions ● DEEPaaS API Support only for synchronous requests ● OpenID Connect support ● Support for standalone service & OpenWhisk action ● Suppor for visual composition of TOSCA templates ● Alien4Cloud PaaS orchestrator support ● Improvements to reach production level ● Virtual Router Virtualized routing over distributed infrastructures ● cloud-info-provider Support for GPU and Infibinand resources ● uDocker Improved support for GPUs and Infiniband ● DEEP-HybridDataCloud 10/10/18 11/14

  12. P 1 st rele DEEP elease: e: ser ervices ices Servic vice Functionalities Preview endpoint Visual application topology Graphical composition of complex application ● composition and topologies https://a4c.ncg.ingrid.pt deployment Deployment through PaaS orchestrator ● Deployment of DEEP Open Catalog ● DEEP as a Service https://vm028.pub.cloud.ifca.es/ components as server-less functions Ready-to-use machine learning and deep ● learning applications, including: ➢ Machine learning frameworks + JupyterLab DEEP Open Catalog https://marketplace.deep-hybrid-datacloud.eu ➢ Machine learning ready to use models ➢ Deep learning ready to use models ➢ BigData analytic tools ● All services are OIDC-ready, following AARC blueprint recommendatjons ● Also work on: TOSCA templates and TOSCA types – Documentatjon and confjguratjon recipes for GPU supoort – Patches to upstream projects (Apache Libcloud, Apache OpenWhisk, OpenStack) – DEEP-HybridDataCloud 10/10/18 12/14

  13. Cont ntact cts Web page: Email: htups://deep-hybrid-datacloud.eu deep-info@listas.csic.es htups://twituer.com/DEEP_eu DEEP-HybridDataCloud 10/10/18 14/14

  14. Thank you Any ny Q Que uestions ions? DEEP-HybridDataCloud This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 777435.

Recommend


More recommend