NeCTAR Melbourne Town Hall

NeCTAR Townhall, 2010-11-26

NeCTAR: National eResearch Collaboration Tools and Resources

$47 million of funding, 2010-2014. Build electronic collaboration infrastructure for national research community. Unimelb is the lead agent.

Aims to enhance research collaboration & outcomes, and to support the connected researcher at the desktop/benchtop. Aims to deploy national research infrastructure and services not otherwise available, in order to enable research collaboration and more rapid outcomes.

Board to approve final project plan, submit to DIISR by Mar 31 2011. Townhall meetings over the next two months.

Consultation paper [PDF] circulated , 60+ responses received, responses available.

Response themes:
* avoid duplicating existing offerings
* needs to be researcher-driven
* questions on how to leverage institutional investments
* need coherent outcomes across nectar
* need to focus on service delivery
* need to establish sustainability

Interim Project plan [PDF] avilable:
NeCTAR is funding four strands of activity. Two are discipline-specific, two are generic and overlaid on the discipline-specific strands.
* Research Tools (discipline-specific, may eventually generalise)
* Virtual labs (resources, not just instruments, available from desktop; emphasis on resources, to prevent them from being applicable to instrument science only).
* Research cloud (general or multi-disciplinary applications and services, plus a framework for using them)
* National server programme (core services, authentication, collaboration, data management services).
NeCTAR will clear up their use of terminology in future communications.

NeCTAR is meant to be serving Research Communities: these are defined as being discipline-based, and range across institutions. e-Research facilitates remote access to shared resources from desktop, in order to enhance collaboration for Research communities (making them Virtual Research communities).

NeCTAR will remain lightweight, to respond to generic and discipline-specific research community needs. Infrastructure is to be built through NeCTAR subprojects. The lead agent UniMelb will subcontract other organisations; some outcomes may be sourced from outside the research community. NeCTAR may start with early adopter groups who already have lots of infrastructure, and NeCTAR may take up existing groupware solutions from these. NeCTAR can only fund infrastructure and not operational services, as it is funded through EIF. Sustainability (as always) is entrusted to the disciplines, NeCTAR will cease at 2014.

Expert panels from across community are to advise the NeCTAR board on allocating subcontracts, as NeCTAR places a premium on transparency. Subcontracts must demonstrate a competitive co-investment model for what NECTAR can't fund: these will take the form of matching funds, likely in-kind, to cover maintenance and support as well as development.
Expert panels will include both researchers, and e-research experts who are familiar with what infrastructure already exists.

There will be a staged model for NeCTAR issuing subcontracts. In 2011 NeCTAR are funding early positive outcomes, in order to give slower-adopting communities more time to develop their proposals. Review of progress and plan for next stage in late 2011.

Research Communities will define the customised solutions they need; these will be delivered through Research Tools & Virtual Labs. Will reserve funds from subcontractors to fund research communities directly, to bring them into Virtual mode.

The considerations for what are the resourcing, scale, timeframe etc of target Virtual Research Communities will inform NECTAR's priorities on what to fund.

NeCTAR is funded to deploy resources for the Cloud nodes, with regard to the Research Cloud, but NeCTAR is not funded to create nodes for the Cloud. NeCTAR will work with existing cloud nodes, e.g. from Research Data Storage Infrastructure (RDSI). Some Research Cloud nodes and RDSI nodes will coexist—but more will be known once the RDSI lead agent has been announced. The consultation responses show a desire for a consistent user experience, which requires a consistent framework for service provision, based on international best practice. (This encompasses virtual machines, data stores access, applications migration, security, licensing, etc.) The framework for the Research Cloud will be developed in parallel with the early projects.

The National Server Program (NSP) will provide core services relevant to all disciplines, e.g. interfaces out of AAF, ANDS, RDSI. The underlying NSP infrastructure will be reliable enough to use as a foundation for more innovative services. The prospect of database hosting has been under much discussion. The National Server Program Allocation Committee is to recommend services for hosting to the NeCTAR board.

Contrast between the National Server Program and the Research Cloud:
* NSP supports core services (EVO, Sharepoint), Research Cloud supports discipline-specific services built on top of the core. (These can include: data analysis, visualisation, collaboration, security, data access, development environments, portals.)
* NSP runs for years, Research Cloud services may only run for minutes.
* NSP provides 24/7 support, Research Cloud provides 9-5 support.
* NSP has strict entry, security, maintenance criteria; Research Cloud less so.

UniMelb is delivering the NSP basic access phase: 50-100 virtual machines, at no charge in 2011, located at UniMelb. This is the first stage of deployment: there will be nodes elsewhere, and Virtual Machine numbers will ramp up.

Many universities are already delivering Virtual Machines, but they can use NeCTAR infrastructure as leverage. Virtual Machine distribution is increasingly used for application release, e.g. with TARDIS.

International exemplars for NeCTAR infrastructure: National Grid Service (UK): Eucalyptus; NASA (US): Open Nebula. NeCTAR will run an expert workshop early next year, inviting international experts and all potential research cloud nodes.

Discussion (from the Twitsphere: #NeCTAR)

* Will the existing ARCS data fabric be maintained? NeCTAR is not able to answer that, since the question is outside NeCTAR's remit. DIISR is in discussions with ARCS on the future of the Data Fabric as well as EVO.

1 comment:

opoudjis said...

Supplemental (from a subsequent town hall):

* Virtual Lab projects will often be a collaboration, but need one lead agency (sub-sub-contracting is needed)
* Virtual Lab project panel will be experts in "priorities for national capability"
* Cloud panels will be "eResearch experts with connections to researchers"
* co-investment example:
* a university was already planning some infrastructure
* NeCTAR funds development and initial installation, the university funds operation and training
* BUT the university must open the infrastructure beyond its own staff
* keen on projects that broaden infrastructure use beyond eResearch experts (e.g HPC experts). Recognise that this requires training rather than software, but can't fund training: hence need for co-investment.