The National Agriculture and Food Research Organization (NARO) has established research facilities nationwide to promote smart agriculture and enhance food self-sufficiency. We spoke with Mr. Eguchi, Part-time Advisor to the Research Center for Agricultural Information Technology, Dr. Katsuragi, Senior Researcher, and Mr. Inoue, Manager of Informatization Promotion (Digital Strategy Department), about the aims and results of data sharing and collaboration using SINET’s network environment. (Interview date: November 7, 2023)

Please tell us about NARO’s role and priority areas in research and development.

Katsuragi:NARO is Japan’s largest agricultural research institute, with a history of over 130 years. It operates 13 research centers and seven research divisions nationwide, with approximately 3,200 researchers and staff. We are engaged in research and development spanning from basic to applied science across diverse fields for the development of agriculture and the food industry. From the perspective of networks and information infrastructure, we are developing research infrastructure encompassing computational resources (supercomputers and cloud computing) and sensor networks. By connecting NARO’s research centers via a network and accumulating and collaborating data, we are advancing research on a nationwide scale. Going forward, we will continue to enhance our research infrastructure capabilities with the aim of achieving smart agriculture and strengthening international competitiveness in the agricultural field. Recently, we have been focusing on research on data-driven agriculture.

What kind of activities does data-driven agriculture involve? Also, tell us about the role of the Research Center for Agricultural Information Technology.

Katsuragi:Data-driven agriculture is an initiative to optimize cultivation technology, work efficiency, and agricultural management based on various data such as weather, soil, crop variety characteristics, yield, and market conditions. In the field of agriculture, many things are done based on the experience and intuition of producers. For example, farmers might delay wheat sowing based on above-average temperatures this year. But relying on intuition alone can produce inconsistent results, and there is an issue in that it is difficult for new farmers to learn. Data-driven agriculture replaces intuition-based decisions with data-driven decision-making processes based on accumulated big data.

The Research Center for Agricultural Information Technology (RCAIT) was established in 2018 to achieve data-driven agriculture using data and artificial intelligence (AI). In addition to the Fundamental Technology Research Division, where RCAIT is located, also houses the Research Center for Agricultural Robotics, the Research Center of Genetic Resources, and the Research Center for Advanced Analysis. RCAIT is working to drive the accumulation and AI analysis of big data on agriculture using research infrastructures such as the AI supercomputer “Shiho” and “NARO Linked Database”, the organization’s integrated database. For example, the system allows us to collect measurement data from individual farms, integrate it into the database, and develop AI models through direct analysis of this data using our supercomputer. Another one of our aims is to return the data and models obtained here to agricultural workplaces by providing them to industry through the WAGRI, the agricultural data collaboration platform that facilitates data sharing across the agricultural sector.

I think the network environment is also important for collecting and linking data. What kind of issues did you face?

Inoue:NARO has 55 research bases, test sites, and seed farms ranging from Hokkaido to Kyushu and Okinawa. The main bases are connected by the Ministry of Agriculture, Forestry and Fisheries Research Network (MAFFIN), but there are also many small bases that needed a cost-effective and reliable network for data integration and supercomputer connectivity as we mentioned earlier. Initially, we tried to connect the whole country through FLET’S (NTT’s fiber-optic internet service), using the NGN network provided by NTT East and West, but there were challenges including separated networks between eastern and western Japan and bandwidth limitations when crossing prefectural boundaries. In addition, since agricultural bases are located in suburban areas, there were issues such as insufficient network performance and vulnerability to power outages caused by lightning strikes.

I see. That’s when we began to focus our attention on SINET.

Inoue:That’s right. SINET maintains at least one node in each prefecture across Japan, so it is possible to connect each base on a prefecture-by-prefecture basis, and when we wanted to create a closed network within a research project in addition to regular Internet access, SINET introduced L2VPN services. It was attractive for us to be able to create purpose-specific isolated virtual networks using multiple VLANs over a single FLET’S connection for multiple experimental networks.
As a result, were able to connect bases (where dedicated line connections were cost-prohibitive) nationwide, uniformly, and at low cost using FLET’S service and SINET.

Tell us about the specific connection configuration.

Inoue:We have converted the connection for our major research bases to 10G, and are using it for data linkage and supercomputer use. However, there was an issue with the cost of connecting SINET to small bases using private lines. To overcome this, instead of private lines, we connected them using FLET’S optical cables. By connecting devices compatible with L2VPN (Ethernet over IP) and devices compatible with VRF-Lite at both ends of the base side and SINET side, and combining VLAN (IEEE802.1q) and VRF-Lite, we were able to connect multiple VPNs between bases via the FLET’S service. By combining FLET’S service and SINET in this way, we were able to achieve both low cost and stable high-speed communication, and an environment in which multiple VPNs can be set up freely between locations.

Katsuragi:In addition, “NARO Linked Database” is deployed on two clouds: NARO Private Cloud and Oracle’s Oracle Cloud Infrastructure (OCI). When accessing the NARO Private Cloud from the outside, we use SINET’s cloud connection service. Specifically, all external access is handled through OCI, which manages authentication, with users then accessing the NARO Private Cloud via SINET’s L2VPN . As a result, safety has been improved, and there is no need to procure new private lines, resulting in lower network costs.

I heard that you are also building a network in preparation for BCP.

Eguchi: We are implementing two key measures. The first part is the establishment of a Business Continuity Planning (BCP) site at a commercial data center in the Sapporo area. In the event of a failure at our headquarters in Tsukuba, we aim to be able to restore operations using the Sapporo data center. Currently we connect to the Internet through MAFFIN, but in the future we are considering L3 connection using SINET’s virtual university LAN.

What are your expectations for the future?

Katsuragi:The journey toward data-driven agriculture has just begun to take shape. Now that we have an infrastructure for collecting research data from each center in one place, we are finally able to share data. We expect this accumulated data to drive advances in both breeding and cultivation technologies.

By connecting to SINET, NARO has secured a stable network environment. By making full use of SINET functionality, we are working to improve communication between sites. Going forward, we will continue to deepen our level of collaboration with SINET and enhance our research infrastructure.

Thank you.