<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta http-equiv="Content-Type" content="text/xhtml; charset=utf-8"> </head> <body> <div style="font-family:sans-serif"><div style="white-space:normal"> Hei Stephan, We, at Uninett, are positive to this proposal and it should provide us feedback from a new user domain. DaaS is a research cluster with 8 GPUs in it (7 Titan X/Xp and 1 P5000) whereas NIRD service platform is a production cluster. Currently NIRD SP lacks GPUs but GPUs (Volta V100) have been ordered and will be available on platform in a month or two. From end user perspective both platform provide a user friendly App Store. So your team can start using DaaS and later on can get access to Volta GPUs on NIRD SP. Regarding access, use the Appstore (<a href="https://appstore.ioudaas.no/" style="color:#3983C4">https://appstore.ioudaas.no/</a>) to deploy <code style="background-color:#F7F7F7; border-radius:3px; margin:0; padding:0 0.4em" bgcolor="#F7F7F7">Deep-learning-tools</code> application. It already has Tensorflow, Pytorch, Keras and few other machine learning framework in it. Documentation is located here <a href="https://appstore.ioudaas.no/docs/" style="color:#3983C4">https://appstore.ioudaas.no/docs/</a> and if your team members would like to have the kubectl level access not just appstore then use docs here <a href="https://docs.ioudaas.no/access/" style="color:#3983C4">https://docs.ioudaas.no/access/</a> Access is controlled using Dataporten, so any user from Norwegian higher education community should be OK also most users from education community in Europe. As Bjørn mentioned I already have a <code style="background-color:#F7F7F7; border-radius:3px; margin:0; padding:0 0.4em" bgcolor="#F7F7F7">nlpr</code> project space. I can create a new group called <code style="background-color:#F7F7F7; border-radius:3px; margin:0; padding:0 0.4em" bgcolor="#F7F7F7">nlpl</code> or rename the current one, please let me know which is preferred. I think, it would be good to have a short (30-60 mins) video meeting next week (Monday anytime between 1000-1200 or Tuesday before lunch time) where I can walk you and your team through the various things to get started and get to know more about your use case as well. -- Gurvinder On 13 Sep 2018, at 12:04, Stephan Oepen wrote: <blockquote style="border-left:2px solid #777; color:#777; margin:0 0 5px; padding-left:5px"> dear colleagues, in our NeIC-funded project NLPL (‘<a href="http://www.nlpl.eu%E2%80%99" style="color:#777">http://www.nlpl.eu’</a>), we are currently running up against limited gpu availability on the Abel and (finnish) Taito clusters. hence, i am wondering whether it would make sense for us to explore usage of the DaaS service platform? we have a handful of technologically reasonably seasoned doctoral fellows who will be eager to gain access to gpus (typically to run TensorFlow or PyTorch). if this sounded like a plausible trial use case to you at this point, i would suggest you set up a namespace (preferably ‘nlpl’ :-) for us and point us at available documentation for users to gain access and get started. all NLPL users (including those from outside norway) have MAS accounts, to access Abel, so i am hoping non-feide users will also be able to connect to DaaS? with thanks in advance, oe </blockquote> </div> </div> </body> </html>