[NLPL Task Force (A)] trial usage on DaaS

[Gurvinder Singh]

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 (https://appstore.ioudaas.no/) to 
deploy `Deep-learning-tools` application. It already has Tensorflow, 
Pytorch, Keras and few other machine learning framework in it. 
Documentation is located here https://appstore.ioudaas.no/docs/ and if 
your team members would like to have the kubectl level access not just 
appstore then use docs here https://docs.ioudaas.no/access/ 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 `nlpr` project space. I can create 
a new group called `nlpl` 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:

> dear colleagues,
>
> in our NeIC-funded project NLPL (‘http://www.nlpl.eu’), 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlpl.eu/archives/infrastructure/attachments/20180913/7ac0fab3/attachment.htm>