[Fiware-fiche-coaching] FW: FIWARE Skype call

["Dave"]

Thanks for the reply. See below for follow up question/clarification

on 20/2/15 6:22 PM, Franck Le Gall <franck.le-gall at eglobalmark.com> wrote:

> Here it is :
>
> · We need to make a financial model for selling a FIWARE based SaaS/PaaS
but
> we do not have any pricing information for the FIWARE nodes (compute,
> memory, storage, bandwidth etc). Understand the software components (GEs)
> are open source and royalty free. But what about the Infrastructure
> services? For now we can make estimates on the IaaS production costs based
> on other providers (Amazon, IBM etc) but looking for direction here from
> FIWARE. When I met the London I3H project they didn't seem to know either.
>
> FIWARE lab is today intended for test & development purpose only. You
should
> not go in production with it. I am not aware of any detailed plan for a
> production grade fiware environment. However, the sustainability plan for
> the fiware community is now being refined: an announcement to support the
> GEs will be made early March at mwc Barcelone. We expect that commercial
> offers for the cloud hosting will follow, either from the community,
either
> from individual organisations, or both, but nothing is set yet.
>

Thanks. This is much the same as I have heard from the I3H UK discussion at
EIT ICT Labs. Challenge we have for FICHe is to make a business
plan/financial model/FIWARE proposal by Feb 28 when the specifics on an
operational FIWARE service (FIWARE Ops Platform Providers) are still
unknown. I think hopefully this can be resolved during the FICHe phase 2.

I am sure will be discussed further on Monday as I know other participants
in Amsterdam bootcamp had same questions.


> · Are there any commercial T+C planned for the FIWARE operations? Service
> level agreements in terms of system uptime, commercial liability,
> indemnification, patent infringement etc
>
> Covered above I think
>
> · FIWARE Lab - Update of Term and Conditions and new user policies.  See
my
> other note.
>
> See first point: not yet intended for production
>
> · Roadmap on Cosmos GE. We have had hard time to get responses from
> Telefonica and the responses we got didn't provide any roadmap. Can FICHe
> team help with an answer here? We are waiting now for 2 weeks for any
> update.
>
> I am investigating this

The other issue with Cosmos is that it can't be made into a dedicated
instance. There is a shared HDFS across all users in the Lab. This makes it
very hard to use with any customer data especially healthcare data sets. See
below from the Cosmos GE catalog

"for the current version of the enabler, there is no automatic deployment
mechanism. It worths mention the current already desployed instance, which
works as global shared instance for both the FI-LAB and FI-PPP memebers, is
not a Cosmos Platform itself, but a shared Hadoop cluster. This is due to
current limitations on the infrastructure, but it will be solved as soon as
more hardware capabilities are added to FI-LAB."

>
> · We need to set-up 4 nodes in the hadoop sense, meaning 4 linux instances
> each with a different public IP.
>
> The availability of public IPs is indeed an issue currently and the new
> policy is expected to solve this. However, could this be made with 1
public
> + 3 private adresses ? (see
>
http://stackoverflow.com/questions/25135183/hadoop-binding-multiple-ip-addre
> sses-to-a-cluster-namenode)
>
> Come back to me as needed

The single public IP doesn't really work as we have a data service that
users can access from their own BI tools.. The platform is such that each
node(instance) in cluster can act as query coordinator. So each cluster node
needs it own IP address that can be access from the outside.

The only workaround for this seems to be limiting the access to REST service
and then have that on a node with public IP and database nodes with private
IP. This is ok for application level access but not for the ad-hoc
analytics/data discovery use cases

>
>
>