Brett Rosolen1, Chris Myers2, David Wilde3, Peter Elford4
1AARNET, Sydney, Australia, Brett.Rosolen@aarnet.edu.au
2AARNET, Melbourne, Australia, Chris.Myers@aarnet.edu.au
3AARNET, Melbourne, Australia, David.Wilde@aarnet.edu.au
4AARNET, Canberra, Australia, Peter.Elford@aarnet.edu.au
The internet is there to share … or is it??
Data intensive research has the potential to generate an incredibly large volume of data. Some of the largest datasets on the planet are in the realm of 100’s of Petabytes! Moving this data for analysis on high performance compute, or simply to collaborate on it, requires high capacity frictionless networks that can perform over very long distances. Large single flows (dubbed “elephant flows”) often dwarf all other traffic and have the potential to consume all available bandwidth.
How can we cater for research that needs almost exclusive use of the network at times without causing a detrimental impact on other users? How do we ensure that the “business as usual” functions of the many are as reliably served as the intensive demands of the few?
Much has already been done to date to enhance Australia’s National Research and Education Network, AARNet, and make this frictionless networking a reality. The national backbone now operates at 100Gbps, with multiple 100Gbps services in place across the Pacific. 100Gbps connections architected for Science at the campus edge (called “Science DMZ”) have been deployed at various sites to improve data transfer while preserving site network security. A suite of tools for data sharing, including Cloudstor and Filesender, have contributed significantly to the daily workflows of the wider research community.
AARNet has been able to demonstrate that a little effort put into border architecture, data transfer tools and data handling workflows can result in extremely large data flows between research infrastructure services and instruments, without impacting on a broader range of users of the entire national network. AARNet is continuing to build new network infrastructure and new platforms and technologies to address this challenge, and to support extreme, unique and evolving research institution requirements.
This talk will identify the science research drivers and subsequent design approach of the network and services, and how the research community can use it to freely move data for better science research outcomes.
Brett’s role as Senior Research Analyst is to ensure that the research community learns to expect more from the significant long term investment in the network and services provided by AARNet, and to understand that research networks worldwide are architected to serve even the most data intensive sciences and collaborations