7. Network Traffic Across the Asia-Pacific R&E Community

Overview

This chapter summarises the scale and growth of research-and-education network traffic carried across the Asi@Connect community between 2013 and 2025, based on figures reported by TEIN NOC and reflected in the e-Compendium portal. Traffic is described by total volume (terabytes carried per year) and peak bandwidth (the maximum throughput reached, taken as the higher of the inbound and outbound peaks); the two directions are not otherwise analysed separately. The main findings are:

• Sustained, accelerating growth. Total volume rose about 3.4-fold since 2014 to roughly 30,658 TB in 2025, while aggregate peak bandwidth climbed even faster — around 16-fold, to about 363 Gbps. Both measures are clearly increasing.
• Traffic is highly concentrated. The three largest networks (JAPAN, CERNET and KREONET) carry roughly 78% of all volume, and the top five about 88%.
• Access now leads by link type. The access layer dominates and continues to grow, reaching the full community total once the Japan trunks are counted as access, while backbone traffic is lower and transit has effectively disappeared since 2016 as direct interconnection matured.
• A pronounced economic divide. Developed and upper-middle-income economies together account for about 97% of 2025 volume, leaving a wide gap to the lower-middle-income, least-developed and regional networks.

The sections that follow develop each of these themes in turn: the overall trend, then traffic volume carried by individual NRENs, by type of link, and by the economic grouping of the economies served.

7.1 Trend analysis

This section examines the network traffic carried by National Research and Education Networks (NRENs) participating in the Asi@Connect community over 2013–2025, drawing on figures published through the Asi@Connect e-Compendium portal for 23 NRENs. Traffic is described by total volume (annual data carried, in terabytes) and by peak bandwidth (the maximum throughput reached, in gigabits per second). The section looks first at the overall trend, then at traffic volume carried by individual NRENs, by type of link, and by the economic grouping of the economies served.

Two conventions apply throughout: a reported value of zero is treated as no traffic for that network and year, and the partial year 2026 is excluded. Figures describe the total traffic carried; incoming and outgoing flows are not analysed separately.

Both measures have risen substantially. Total volume grew from roughly 5,446 TB in 2013 to about 30,658 TB in 2025 — around 3.4-fold since 2014, a compound rate near 12% per year — while aggregate peak bandwidth climbed even faster, reaching about 363 Gbps in 2025. The volume trend and the aggregate peak-bandwidth overlay are shown together below.

Growth was not uniform. Volume was comparatively flat from 2013 to 2018, stepped up sharply in 2019, and has since settled into a higher, somewhat volatile band, reaching a high of 30,658 TB in 2025 (with 2023 close behind at 30,124 TB). Peak bandwidth followed a broadly similar but steeper path, reflecting heavier instantaneous demand on the busiest links.

7.1.1 Peak bandwidth by network

Peak demand is dominated by a few networks. In 2025 JAPAN reached about 183 Gbps — far above the rest — and the three highest-peaking NRENs together accounted for roughly 81% of all peak bandwidth. The chart below ranks each network by its 2025 peak (logarithmic scale).

Note: SG-SG traffic carried both SingAREN and SG-EU-LON traffic from July 2019 to November 2025.

Viewed over time, the leading networks’ peaks have escalated dramatically. Every one advanced severalfold between 2013 and 2025 — JAPAN around fifty-fold — several from a near-standing start, as the chart below shows.

Note: SG-SG traffic carried both SingAREN and SG-EU-LON traffic from July 2019 to November 2025.

7.2 Traffic volume carried by individual NRENs

Total traffic volume is highly concentrated among a small number of networks. In 2025 the three largest — JAPAN, CERNET and KREONET — together carried roughly 78% of all traffic, and the top five about 88%. The chart below ranks each network by its 2025 volume (logarithmic scale).

The same concentration is evident cumulatively: a handful of networks account for the bulk of regional traffic, with the cumulative share crossing 80% within the first four networks.

7.2.1 Per-NREN trends

The panels below trace each NREN’s total annual traffic volume across 2013–2025, arranged alphabetically; the figure beside each name is its 2025 total. They show a range of trajectories: steady climbers such as JAPAN, CERNET and KREONET; networks that peaked mid-decade and then receded; and several that recorded little or no traffic in recent years. JAPAN aggregates the SG–JP and HK–JP links reclassified as access; DrukREN is shown with no data because it routes all of its traffic through NKN, which did not report statistics.

7.3 Traffic volume carried by type of link

Traffic can also be examined by the segment of the network it traverses. The statistics distinguish five link types: access links (edge connections between member institutions and their NREN), backbone links (the core inter-node spine), gateway links (interconnections to external networks), transit links (traffic passing through on the way elsewhere), and other links (a residual category carrying measurable traffic mainly before 2020). The table summarises 2025 volume and peak bandwidth for the four types still active that year.

By volume the access layer now carries the most traffic at roughly 30,658 TB — equal to the community total once the Japan trunks are counted as access — while the backbone handled about 8,850 TB, gateway about 5,819 TB, and transit and other links effectively none. The chart below shows how each link type has grown over the period.

Two structural shifts stand out: the access layer now dominates and continues to grow, while transit collapsed after 2015 — from a peak above 3,000 TB to virtually nothing since — consistent with a move away from third-party transit toward direct interconnection, a common sign of a maturing regional fabric. The access layer also records the highest peak bandwidth (about 221 Gbps), reflecting the high-capacity Japan trunks now counted there.

7.4 Traffic volume carried by each economy

Grouping NRENs by the economic classification of the economies they serve exposes a structural divide across five bands: developed (DC), upper-middle-income (UMIC), lower-middle-income (LMIC), least-developed (LDC), and regional or special-purpose networks (Others).

In 2025 the developed-economy group carried about 80% of all volume and the upper-middle-income group about 18% — together roughly 97%. The lower-middle-income and least-developed groups together accounted for under 3% of 2025 volume, despite comprising twelve of the 23 member networks. The chart below shows how each group’s share of total traffic has shifted over time.

The developed-economy share rose from about 55% in 2013 to roughly 80% in 2025, while the upper-middle-income group eased from about 22% to 18%. The lower-middle-income share fell from around 21% to about 2% as several of its networks recorded little or no traffic in later years. Because the two smallest groups are barely visible above, the next chart shows their volumes directly on a logarithmic scale.

In absolute terms the regional total is carried overwhelmingly by the developed and upper-middle-income groups, as the stacked and ranked views below confirm.

7.5 Traffic composition and upstream capacity

Beyond how much traffic flows, the survey records the maximum upstream bandwidth each NREN consumes by category, from which its traffic-type mix can also be read: a category counts as carried when its reported upstream bandwidth is greater than zero. On that basis in 2025, research-and-education (R&E) and internet-exchange traffic are the most widespread — carried by 15 of the 19 reporting NRENs each — followed by commodity or general-internet traffic (14) and content-delivery (CDN) traffic (9). Eight NRENs run the full four-way mix: REANNZ, HARNET, PREGINET, SingAREN, LEARN, ThaiREN, BdREN and ErdemNET. Research-and-education upstream bandwidth is led by New Zealand's REANNZ (about 220,000 Mbps) and Hong Kong's HARNET (about 200,000 Mbps), with Japan's MAFFIN next at 100,000 Mbps; most other NRENs report R&E upstream in the low thousands of Mbps or below. CERNET, LERNET and NKN did not report any upstream bandwidth in 2025.

Summary

Across 2013–2025 the Asi@Connect community’s traffic grew about 3.4-fold in volume to roughly 30,658 TB, while aggregate peak bandwidth rose faster still, reaching around 363 Gbps — both measures point to steadily intensifying demand. Volume is heavily concentrated among a handful of high-capacity networks; by link type the access layer dominates while transit has receded; and a clear economic gradient leaves developed and upper-middle-income members carrying the large majority of traffic. Taken together, these trends describe a maturing ecosystem in which a few well-resourced NRENs increasingly anchor regional research-and-education traffic, while a long tail of smaller networks grows from a much lower base.

7.6 Traffic volume by NREN (data appendix)

Total annual traffic volume (inbound plus outbound, in terabytes) for each participating NREN at each reporting year. A value of 0 indicates no traffic was recorded for that year; “<1” indicates a value below 1 TB.

Source: Asi@Connect e-Compendium portal. Zero values treated as no traffic; 2026 excluded as a partial year. JAPAN aggregates the SG–JP and HK–JP links reclassified as access.

7.7 Network links by type

Section 7.3 grouped traffic by type of link. For reference, the table below lists the individual Asi@Connect links that make up each type, with the provisioned capacity of each and remarks noting status changes where applicable (for example, terminated or upgraded links). The classification follows the network’s operational link naming.

Capacities are provisioned link rates (M = Mbit/s, G = Gbit/s). 60 links in total: 46 access, 1 transit, 3 backbone, 7 gateway, 3 other.