Speakers
Marcus Hardt
(KIT-G)
Patrick Fuhrmann
(DESY)
Description
The INDIGO DataCloud project is set out to develop a data/computing
platform targeting scientific communities, to enhance usefulness of
existing e- infrastructures [1]. The developments in the storage are
focus on two levels. On the IaaS level, QoS in storage will be adressed,
by implementing a standardised extension to the CDMI standard, which
enables management of storage quality e.g. access-latency,
retention-policy, migration-strategy or data-lifecycle. This is closely
related with intelligent identity management to harmonise access via
different protocols, such as gridFTP, sftp and CDMI. This will allow to
use CDMI to manage QoS of data that is accessible via gridFTP or sftp.
On the PaaS level, INDIGO DataCloud will provide flexible data federation
functionality, enabling users to transparently store and access their data
between heterogeneous infrastructures. DataCloud will provide unified
API's for data management based on state of the art standards, allowing
both users and application developers to easily integrate DataCloud high
level data management functionality into their use cases. One of the key
features of this solution will be optimization of data access in various
scenarios, which will include automatic pre-staging, maximum bandwidth
usage via parallel transfers and enabling instant access to remote data
through streaming. Furthermore, the layer will provide information to
Cloud orchestration services allowing placing the computations in the
sites where the data is already staged, or where it can be delivered
efficiently.
[1]https://www.indigo-datacloud.eu/
Primary authors
Dr
Giacinto Donvito
(INFN)
Dr
Lukasz Dutka
(CYFRONET)
Marcus Hardt
(KIT-G)
Patrick Fuhrmann
(DESY)
