China turns on a vast experimental network it says is an heir to ARPANET

Chinese authorities on Thursday certified the China Environment for Network Innovation (CENI), a vast research network that Beijing hopes will propel the country to the forefront of networking research.

As reported in Chinese state media, tests of the network saw it shift 72 terabytes of data in 1.6 hours, across a distance of around 1,000 km between a radio telescope in Guizhou province and a university in Hubei. We think that’s almost 100 Gbit/s, an impressive feat for a sustained long-distance data transfer even if it took place in a controlled environment.

CENI took over a decade to build and now links 40 Chinese cities with over 55,000 kilometers of optic fiber. State media says it can support 128 heterogeneous networks and 4,096 parallel heterogeneous service tests.

China often lets its techies describe the nation’s technological triumphs at international conferences, and in November one of them – Bingqing Wu of the Jiangsu Future Network Innovation Institute – delivered a talk about CENI at the HEPiX Forum, a conference at which high-energy physics types talk tech.

His presentation about CENI is an eye-opener as it compares China’s network to ARPANET and the Global Environment for Network Innovations (GENI) – US research networks used to do much of the formative work that gave us the internet and so much more. Both ARPANET and GENI have been decommissioned, although many nations and institutions operate more modern research networks – a class of facility referred to as a “national research and education network.”

Here’s another goal for CENI mentioned in the presentation:

The HEPiX presentation also mentions one of the goals for CENI is to develop networking innovations “5-10 years ahead of the industry,” and that Chinese tech giants Huawei and Baidu have already used the network to test their tech. The latter company has apparently already used CENI to vastly increase the efficiency of corpus acquisition when moving data for AI inferencing and training workloads.

That’s a notable mention as China is trying to build a homebrew AI stack. CENI shows it already has a very handy tool to advance that effort. ®

READ MORE HERE