Global Lambda Integrated Facility

GLIF goes walkabout Down Under

3 October 2017 -- The 17th Annual Global LambaGrid Workshop was held on 25-27 September 2017 at the University of Sydney Camperdown, Sydney, Australia. 70 participants from 16 countries came together to hear about the recent developments in Australia and the Pacific region that are supporting high-bandwidth research applications such the Square Kilometre Array, and to discuss how global R&E networks should evolve over the next 5 years. Thanks go to AARNet for hosting and sponsoring the workshop.

The keynote was provided by Dr. Shaun Amy (CSIRO) who discussed the data challenges posed by the Square Kilometre Array (SKA). This is a multinational radio telescope initiative that is primarily sited in the Southern Hemisphere because it faces the galactic centre and is therefore better placed for observations. This requires provision of long-haul, high bandwidth networks to very remote locations, as well as establishing high-performance computing and storage centres for handling and processing the data; expected to be in the order of 300 petabytes per year. This will in turn typically require 100-200 Gb/s connections from each telescope to localised processing facilities, and then onwards to Europe and North America.

Garret Yoshimi (University of Hawaii) then discussed the initiatives expanding R&E networking in the Pacific. This constituted a huge area of scattered islands, and whilst Hawaii is fortunate to have multiple 100 Gb/s circuits to the US and Australia/NZ, much of the rest of the Pacific has limited connectivity. The Pacific Islands Research and Education Network (PIREN) funded by the US National Science Foundation is therefore establishing 100 Gb/s links to Guam and onwards to Singapore (SingAREN), with possible future links from Guam to Japan and Australia facilitated by new cable builds in the region. This connectivity is needed to support the Mauna Kea observatories, SKA, coral reef monitoring networks, and research visualisation amongst other applications.

Rommel Hidalgo added some further information about Guam Open Research & Education Exchange (GOREX) that has been established by the University of Hawaii, AARNet, CENIC and Pacific Wave, with the University of Guam hosting and operating the facility, and NSRC providing campus network design capacity building. This aims to take advantage of the strategic location of Guam, and to improve access to global R&E networks from Pacific Island nations whilst improving resilience in the region.

The participants then heard about other specific applications taking advantage of the GLIF infrastructure, including from David Abramson (University of Queensland) on the MeDiCI infrastructure that extends data caching across the Pacific, and from Joe Mambretti (Northwestern University) on the Global Research Platform which is using SDN to create a distributed scientific compute and storage cloud. This was followed by an interesting talk from Tomohiro Kudoh (University of Tokyo) on how energy consumption of conventional communication networks is becoming difficult to meet, despite the improvements in the energy efficiency of equipment. One solution may be Dynamic Optical Path Networking offering all-optical switching, and the VICTORIES project, supported by the Japanese Ministry of Education, Culture and Sports in collaboration with AIST, has developed a testbed offering much lower energy consumption combined with lower latency and the potential to support quantum networking in the future.

The following afternoon and morning sessions were devoted to technical discussions chaired by Lars Fischer (NORDUnet) and David Wilde (AARNet). The main focus was the current challenges faced by R&E networks and what the directions need to be for the next 3-5 years. Some of the issues identified were that networks are increasingly software driven, and software systems and software development are becoming as important as network engineering whilst (N)RENs traditionally have limited experience in developing and testing production code. Similarly, more services were moving into the cloud, and whilst this hides the complexity from users, it makes managing networks more complex and will require more sophisticated measurement and monitoring. More generally, many (N)RENs could not continue to justify higher bandwidth networks on a handful of big data research projects alone, and also needed to ensure good access to compute and storage clusters for the smaller research projects too.

Other developments were the establishment of a new GOLE called MOXY in Montreal earlier in 2017, that has 100 Gb/s connections to CANARIE, StarLight, MAN LAN and NetherLight. The MEICAN AutoGOLE pilot has also developed a graphical front-end for the AutoGOLE system that allows multi-domain lightpath services to be provisioned dynamically, and plans to trial this in production networks during Q3 of 2017. Finally, an initiative to document GOLE attributes was started by GLIF that aims to list services, capabilities and policies for each GOLE, in order to compare operational capabilities and provide guidance to users.

Following the end of the first day, several demonstrations utilising the 400Gbps of network capacity AARNet delivered to the venue. This included Ciena's bandwidth demand prediction mechanism, CESNET's SDN-controlled live optical path protection (actually controlled by a model train breaking a connection), Northwestern University's Software-Defined Wide-Area Networks (SD-WANs), SURFnet's demo of a DTN from NetherLight to StarLight, and an AutoGOLE Task Force demo of the MEICAN control plane.

The Governance Working Group chaired by Jim Ghadbane (CANARIE) was held the day before the workshop. The most important announcement was the launch of the new GLIF maps, whilst the budget for 2017 was approved and again keeps the sponsorship tiers at existing levels. Several options for the future hosting of the GLIF Secretariat were discussed, and it was agreed to further investigate a couple of the options. Jonah Keough (Pacific Northwest GigaPoP) also volunteered to join the Programme Committee for GLIF 2018, which otherwise remains unchanged.

The other meeting held the day before the workshop was the GLIF Americas Working Group (chaired by Joe Mambretti, StarLight International/National Communications Exchange). This discussed developments and requirements in North and South American R&E networking.

The closing plenary session had to be extended because of the number and quality of the presentations submitted, and kicked off with JP Velders (University of Amsterdam) discussing trans-oceanic performance engineering and monitoring. Keeping with the monitoring theme, Jeronimo Bezerra presented AmLight’s SDN Looking Glass which offers centralised SDN monitoring, whilst Jerry Sobieski (NORDUnet) concluded the morning by asking what happened with the NSI stuff?

Jamie Curtis (REANNZ) provided a very informative talk on the engineering challenges of building the fibre system used for REANNZ's new international network, followed by Xing Li (CERNET) who discussed the Chinese experiences with rolling out a DWDM 100 Gb/s system to support their different IPv4 and IPv6 overlay networks. Guy Roberts rounded off the session by highlighting the ongoing dis-aggregation of optical network platforms and the move towards lower-cost Data Center Interconnect (DCI) equipment. The Telecom Infra Project (TIP) initiated by Facebook aims to achieve an open market for networking equipment by building modular commodity units that will substantially bring down the cost of the hardware needed to support the growth of the content provision market. GÉANT has been testing DCI boxes from several vendors, and is looking to choose a solution in order to more cheaply and effectively upgrade its network to meet growing demand.

Last but very much not least was a panel session on Services, Performance Engineering and Monitoring across Oceans. This was moderated by Rodney Wilson (Ciena) and featured Xing Li (CERNET), Joe Mambretti (Northwestern University), JP Velders (University of Amsterdam) & Garret Yoshimi (University of Hawaii).

The workshop concluded with a closing address from GLIF Co-Chair David Wilde who summarised the proceedings and thanked the participants for contributing to the vibrant discussions over the past three days. The dates and venue for the 18th Annual Global LambdaGrid Workshop will be 20-21 September 2018 in Helsingør (Elsinore), Denmark, co-located with the NORDUnet Conference 2018.

We look forward to seeing you in Denmark next year!

The proceedings of the workshop are available at http://www.glif.is/meetings/2018/

About GLIF -- The Global Lambda Integrated Facility (GLIF) is an international virtual organisation of NRENs, consortia and institutions that promotes lambda networking. GLIF provides lambdas internationally as an integrated facility to support data-intensive scientific research, and supports middleware development for lambda networking. It brings together some of the world's premier networking engineers to develop an international infrastructure by identifying equipment, connection requirements, and necessary engineering functions and services. More information is available on the GLIF website at http://www.glif.is/