Conclusions and Outlook - Science Gateways for Distributed Computing Infrastructures - page 278

5. Enhanced cloud integration. Cloud integration will play a more and more
important role in future gateway operations. Therefore both the CloudBroker
based integration and the direct integration will be signi

cantly improved and
extended in the future. For example, currently services cannot be launched in
clouds via the CloudBroker integration mechanism. This issue will be solved
within the CloudSME project by winter 2014.
6. Installing and con guring WS-PGRADE/gUSE-based gateways. During the

SCI-BUS project the installation and con guration methods have been

signi cantly improved. The installation and con guration wizardz released in

Spring 2012 have signi cantly simpli ed these procedures. However,

con guring the collaboration of the gateway with some of the DCIs still can

cause problems for the gateway operators if they do not have an understanding
of the connected DCI. Therefore troubleshooting when the con

guration of the
gateway with a DCI does not work correctly became a major issue. In order to
ease troubleshooting, the gUSE team continuously extends and rationalizes the
logging mechanism used in WS-PGRADE/gUSE.
7. Deployment of WS-PGRADE/gUSE-based gateways in clouds. Another

possibility to facilitate the installation and con guration of WS-PGRADE/

gUSE gateways is the deployment of the gateway in clouds. Recently a WS-
PGRADE/gUSE gateway image has been published in Amazon that enables
anyone to deploy the gateway in Amazon in a fully automatic way. The gateway
is created on-demand and stays alive as long as the user needs it. The current
gateway image is con gured to use the local submitter option, i.e., it will

execute the jobs on the cloud resource where the gateway is deployed. This
solution will soon be available in academic clouds as well. This approach raises
a new gateway usage model that discussed in the next section.

Future Models of Using Gateways

The current model of using gateways is based on 24/7 gateway services. This means
that a user community establishes a server that runs the scienti c gateway and

members of the community can access the gateway service at any time and from
any continent.
This model was further generalized by the StarNet gateway alliance model
(Chap. 18 ) . The idea of this model is that for large communities that have several
subcommunities not only a single gateway is established; rather a set of gateways
serving the different subcommunities are established. These gateways are very
similar using a core image containing the basic common functionalities of these gate-
ways. This common image aids in quick installation and con guration when these

gateways are set up. However, since the different subcommunities also want some
speci

c features, these gateways are not perfectly uniform. They are typically